{"schema_version":1,"scope":"All published catalogue resources","count":270,"resources":[{"slug":"a-lab","title":"A-Lab: an autonomous laboratory for the accelerated synthesis of inorganic materials","summary":"An autonomous laboratory for solid-state synthesis of inorganic powders that combines computations, literature-mined data, machine learning and robotics. Over 17 days of continuous operation the authors report it realized 41 novel compounds from a set of 58 targets, including oxides and phosphates.","resource_type":"use-case","disciplines":["materials-science"],"tasks":["data-analysis","experiment-execution","materials-discovery"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":"L4","workflow_note":"Iterative synthesis planning, robotic execution and interpretation with active-learning feedback within predefined targets. Humans provide targets and infrastructure, oversee safety and validate phase identification; autonomous operation does not establish correctness.","source_checked_date":"2026-10-15","limitations":"Reported performance is author-reported and was debated publicly after publication; a 2025 Author Correction was issued (see sources). Scope is solid-state synthesis of inorganic powders, and the pipeline depends on specific robotics and on large ab initio datasets. Not independently reproduced.","adoption_requirements":"Robotic solid-state synthesis and characterization infrastructure; compute for ab initio phase-stability analysis; access to Materials Project data; expertise to triage failed syntheses rather than accept robot output blindly.","updated_at":"2026-10-15T17:25:09Z","url":"https://research.wusong.net/cases/a-lab","sources":[{"label":"Szymanski et al., An autonomous laboratory for the accelerated synthesis of inorganic materials, Nature (2023)","url":"https://doi.org/10.1038/s41586-023-06734-w","kind":"doi","is_primary":1},{"label":"Author Correction (Nature, 2025)","url":"https://doi.org/10.1038/s41586-025-09992-y","kind":"correction","is_primary":0}]},{"slug":"academic-writer-yodem","title":"Academic Writer: source-linked humanities drafting","summary":"A documented Claude Code plugin for Hebrew/English humanities articles connects source/thesis selection, approved outlines, drafting, citation checking and Word-comment revision.","resource_type":"agent-skill","disciplines":["humanities"],"tasks":["knowledge-synthesis","literature-research","scientific-writing"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Intended L2: researcher-selected sources and thesis, approved section claims/outline, model drafting and citation checks lead to Word output and researcher-confirmed revision. Checks/retries are documented, not independently verified. Fixed drafting stages do not establish adaptive L3 or autonomous science.","source_checked_date":"2026-10-19","limitations":"Selected README writing workflow, setup/prerequisites, privacy/library caveats and contributor/testing context read; linked full guide and code not audited. No claim that every citation is actually correct or all controls enforced. README describes conversation/source passages sent to Anthropic, drafting/extraction and other text sent to Google with Gemini configuration, and optional source/profile storage in Trove; current provider terms not checked. Quiet library test/explicit AI evaluation and logs are documented features, not experiments run here. No real manuscripts, keys, source libraries or private profiles accessed. AI editorial review only; no source code, models or installation executed, no production/status/publication changes.","adoption_requirements":"README names Claude Code, a Gemini API key for full drafting/research functions, Python and Word-compatible editing, plus optional source libraries. Plan/cost/free-tier and setup claims are unverified current documentation; no installation or account access performed.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/academic-writer-yodem","sources":[{"label":"yodem/academic-writer primary-readme","url":"https://raw.githubusercontent.com/yodem/academic-writer/3d485019121394b8461882b3b9c5cc07514339b4/README.md","kind":"code-repository","is_primary":1}]},{"slug":"ada-self-driving-laboratory","title":"Ada (self-driving laboratory)","summary":"Self-driving laboratory that defined the Pareto front of conductivities and processing temperatures for palladium films via combustion synthesis, including conditions enabling coating of commodity plastics.[9]","resource_type":"use-case","disciplines":["chemistry","materials-science"],"tasks":["experiment-execution","materials-discovery","optimization"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans set objectives and validate results.","source_checked_date":"2026-10-16","limitations":"Author-reported outcomes; not independently reproduced. The published version was checked at record/abstract level; full text not inspected.","adoption_requirements":"Autonomous-laboratory hardware and software stack; not packaged for third-party adoption in the checked sources.","updated_at":"2026-10-17T08:00:32Z","url":"https://research.wusong.net/cases/ada-self-driving-laboratory","sources":[{"label":"Advancing the Pareto front for thin-film materials using a self-driving laboratory - arXiv:2106.08899","url":"https://arxiv.org/abs/2106.08899","kind":"arxiv","is_primary":1}]},{"slug":"aeneas","title":"Aeneas","summary":"Aeneas helps historians retrieve parallel Latin inscriptions and propose restorations, dates and geographical attributions.","resource_type":"tool","disciplines":["humanities"],"tasks":["archival-research","data-analysis"],"evidence_level":"institutional-report","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Reviewed the team announcement, not the complete Nature study or evaluation data. Restoration and attribution remain hypotheses requiring epigraphic expertise; no performance figures are adopted.","adoption_requirements":"Inscription text or images, historical expertise, and rights to research material.","updated_at":"2026-10-16T14:58:30Z","url":"https://research.wusong.net/cases/aeneas","sources":[{"label":"Aeneas transforms how historians connect the past - Google DeepMind","url":"https://deepmind.google/blog/aeneas-transforms-how-historians-connect-the-past","kind":"institutional-report","is_primary":1}]},{"slug":"agentic-ai-literature-review-bitik","title":"Academic literature scanning agent (Bitik)","summary":"A documented Python/LangChain workflow uses a researcher topic to retrieve papers/datasets, suggest literature gaps and save a structured Markdown report.","resource_type":"tool","disciplines":["multidisciplinary"],"tasks":["knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Scoped documented L2: topic/model selection connects literature and dataset retrieval, gap suggestions and a Markdown report. Autonomous branding and tool references do not establish adaptive planning or L4. Human source/output validation is not documented.","source_checked_date":"2026-10-19","limitations":"Documentation only; no repository code, model calls, installation, data processing or experiment executed. Primary API/raw source retrieval does not validate implementation, security, performance or scientific outcomes. Turkish README reviewed; no code/linkedAPI/model run. Lastfiveyears and citation sorting are documented intended scope, not exhaustive/current literature coverage or study-quality ranking. Dataset/gap/novelty suggestions and Markdown reports are not verified evidence. API availability/pricing, language generality, citation accuracy and human validation remain untested.","adoption_requirements":"README specifiesPython3.10+,LangChain,OpenAI-compatible model configuration andSemanticScholarAPI. Credentials/model terms and provider availability must be checked independently; no setup performed.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/agentic-ai-literature-review-bitik","sources":[{"label":"furkan-bitik-dev/Agentic-AI-Literature-Review primary-readme.json","url":"https://raw.githubusercontent.com/furkan-bitik-dev/Agentic-AI-Literature-Review/e610d095f460890ea2a9724f9f37b5b46d0a9bb3/README.md","kind":"code-repository","is_primary":1}]},{"slug":"ai-assisted-history-of-technology-research","title":"AI-assisted history-of-technology research workflow","summary":"An author-described historical research workflow that decomposes a contemporary AI application into technical components, searches for foundational research and uses generative AI for iterative discussion, transcription summaries and writing support.","resource_type":"workflow-process","disciplines":["computer-science","humanities"],"tasks":["knowledge-synthesis","literature-research","scientific-writing"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level workflow automation intentionally unspecified. A reusable component/process reference is not a measured autonomous workflow; researcher oversight remains necessary.","source_checked_date":"2026-10-17","limitations":"ArXiv abstract plus introduction, research questions/objectives and section2 method excerpt reviewed; sections3 onward, historical references and factual chronology were not audited. No comparative benefit, comprehensive search, historical accuracy or child-safety finding is established. No software, speech recordings or outcome was reproduced. No journal publication was identified in the inspected primary record.","adoption_requirements":"Editorial, untested: an academic-search workflow and optional generative-AI/transcription access would be needed for adaptation. Check source references and rights yourself; do not upload confidential recordings or archival material without permission. No exact model version, prompt recipe or cost is established.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/ai-assisted-history-of-technology-research","sources":[{"label":"Why are we living the age of AI applications right now? The long innovation path from AI's birth to a child's bedtime magic","url":"https://arxiv.org/abs/2501.06929","kind":"arxiv","is_primary":1},{"label":"Primary introduction and research-method excerpt","url":"https://arxiv.org/html/2501.06929v1","kind":"arxiv","is_primary":1}]},{"slug":"ai-auto-research-four-phase-roadmap","title":"AI for Auto-Research: four-phase research lifecycle roadmap","summary":"A conceptual roadmap organising AI-assisted research into Creation, Writing, Validation and Dissemination, with an emphasis on human-governed collaboration rather than unchecked autonomy.","resource_type":"tutorial-template","disciplines":["multidisciplinary"],"tasks":["literature-research","scientific-writing","workflow-orchestration"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation unspecified; researchers select the task, assess data rights and inspect outputs. Source review is not an executed workflow.","source_checked_date":"2026-10-17","limitations":"Abstract-only arXiv v2 review. The project site, tool inventory, benchmark suite and detailed playbook were not inspected. The abstract's cost, acceptance and model-reliability comparisons are not adopted as independently verified conclusions; no quantitative performance outcome is claimed.","adoption_requirements":"Editorial, untested: map a real project onto the four phases, assign researcher oversight, and separately assess model access, data confidentiality, licensing and validation requirements. No specific tool is endorsed.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/ai-auto-research-four-phase-roadmap","sources":[{"label":"AI for Auto-Research: Roadmap & User Guide","url":"https://arxiv.org/abs/2605.18661","kind":"arxiv","is_primary":1}]},{"slug":"ai-bubble-stress-engine","title":"AI Bubble Stress Engine","summary":"Author-described financial research project prepared with Claude assistance that scores stress in the AI investment cycle from public evidence, records counter-evidence and states conditions that would change its verdict.","resource_type":"use-case","disciplines":["business","economics"],"tasks":["data-analysis"],"evidence_level":"repository-documented","outcome_verification":"author-reported","workflow_level":"L1","workflow_note":"Researcher-directed AI assistance; humans choose objectives, inspect sources and approve outputs. L1 denotes disclosed assistance, not autonomous collection or scoring.","source_checked_date":"2026-10-17","limitations":"README-only review; scoring code, underlying indicators and tests were not executed or audited. README current block says 18 indicators while the known-weakness section says 17, and robustness prose retains 58 while the current block shows 57: version drift is unresolved. Judgment-heavy inputs, unconfirmed private-company figures and no historical backtesting limit inference. Not investment advice.","adoption_requirements":"Python 3 according to README; analyst judgment and independently checked public financial sources. Verify dataset permissions and private-company claims. Retain researcher oversight and do not use the stress index as a validated trading signal.","updated_at":"2026-10-17T17:46:50Z","url":"https://research.wusong.net/cases/ai-bubble-stress-engine","sources":[{"label":"AI Bubble Stress Engine — primary README","url":"https://raw.githubusercontent.com/ELMAJ22/ai-bubble-stress-engine/HEAD/README.md","kind":"code-repository","is_primary":1}]},{"slug":"ai-ethics-discourse-nlp-workflow","title":"AI ethics discourse: mixed qualitative and NLP analysis workflow","summary":"A research workflow combining qualitative coding with NLP topic modelling to study changes in OpenAI ethical discourse across public and academic-facing documentation.","resource_type":"workflow-process","disciplines":["humanities","social-science"],"tasks":["data-analysis","knowledge-synthesis"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation unspecified; researchers select the task, assess data rights and inspect outputs. Source review is not an executed workflow.","source_checked_date":"2026-10-17","limitations":"Abstract-only review of arXiv v1. The linked code repository and related journal DOI were identified but not inspected. NLP model, corpus coverage, coding reliability, preprocessing and implementation requirements are unknown. No discourse findings or efficacy estimates are adopted as verified outcomes.","adoption_requirements":"Editorial, untested: obtain the full methods and linked code, confirm document reuse rights and corpus coverage, and plan human checking of both topic interpretations and qualitative codes.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/ai-ethics-discourse-nlp-workflow","sources":[{"label":"Competing Visions of Ethical AI: A Case Study of OpenAI","url":"https://arxiv.org/abs/2601.16513","kind":"arxiv","is_primary":1}]},{"slug":"ai-for-ee-research-literature-guidance","title":"AI for EE Research: supervised literature and verification guidance","summary":"Selected textbook guidance for electrical-engineering researchers connects AI-assisted literature searching, evidence extraction and synthesis with explicit researcher checks.","resource_type":"tutorial-template","disciplines":["computer-science","engineering"],"tasks":["knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Intended L2 instructional guidance for planning a literature search, organising and synthesising evidence, and checking outputs and reproducibility. Researcher judgement remains central. These are source-prescribed steps, not verified implementation, scientific accuracy or autonomous research.","source_checked_date":"2026-10-19","limitations":"Book index and selected literature-search/synthesis and verification/reproducibility chapters only. The instructional material is AI-generated; it is not independently verified expert guidance. The whole book, linked studies, examples and code were not audited or executed. Source and output checks are recommendations, not demonstrated enforcement. Moving HEAD documents are unpinned and non-atomic.","adoption_requirements":"Untested suggestion: verify each cited primary source, adapt the guidance to a defined research question and check outputs independently before reuse.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/ai-for-ee-research-literature-guidance","sources":[{"label":"Book index","url":"https://raw.githubusercontent.com/dmccreary/ai-for-ee-research/HEAD/docs/index.md","kind":"project-documentation","is_primary":1},{"label":"Literature search and synthesis chapter","url":"https://raw.githubusercontent.com/dmccreary/ai-for-ee-research/HEAD/docs/chapters/11-literature-search-synthesis/index.md","kind":"project-documentation","is_primary":1},{"label":"Verification and reproducibility chapter","url":"https://raw.githubusercontent.com/dmccreary/ai-for-ee-research/HEAD/docs/chapters/17-verification-reproducibility/index.md","kind":"project-documentation","is_primary":1}]},{"slug":"ai-literature-review-suite-tovar","title":"AI Literature Review Suite: paper-described search and synthesis tool","summary":"A primary preprint abstract describes an AI literature-review suite combining semantic retrieval, PDF organisation, extraction and LLM summarisation.","resource_type":"tool","disciplines":["multidisciplinary"],"tasks":["knowledge-synthesis","literature-research"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level assigned. Semantic search, PDF retrieval and organisation, extraction, embeddings, summaries and PDF interaction are described, but stage control and researcher output validation are not established. A GUI alone does not demonstrate oversight.","source_checked_date":"2026-10-19","limitations":"Official preprint abstract only. No full paper, code, GUI, corpus, citation accuracy, licence or provider handling was audited. Claims of robustness, optimisation and automation remain author descriptions. Same-named repositories are not assumed to be canonical distributions. No empirical benefit is established.","adoption_requirements":"Untested suggestion: inspect the original paper and clarify distribution provenance, PDF rights, provider handling and researcher source checks before a trial.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/ai-literature-review-suite-tovar","sources":[{"label":"Primary preprint abstract","url":"https://arxiv.org/abs/2308.02443","kind":"arxiv","is_primary":1}]},{"slug":"ai-ml-questions-santana","title":"AI/ML Questions: AI-assisted item development and measurement","summary":"A documented AI/ML question-bank project connects AI-assisted drafting/review with intended human approval and item-level measurement/revision guidance.","resource_type":"tutorial-template","disciplines":["computer-science","education"],"tasks":["data-analysis","knowledge-synthesis"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Intended L2 for AI-assisted item drafting, deterministic checks, second-model review and intended human approval, followed by documented item-statistics/revision guidance. Current alpha drafts have not all received human review. No learner-data audit, representative sampling or learning/psychometric effectiveness is established.","source_checked_date":"2026-10-19","limitations":"Full README and statistics documentation read; no question contents, individual responses, database, app interaction or source code inspected/executed. Anonymous/no-cookie/no-IP data handling, counts and calculations are source claims, not audited privacy/results. Statistics guide warns self-selected users, per-browser first attempts, study-mode/order/attrition effects, version filtering and possible fabricated responses. Intended human approval must not be reported as already completed. No population generalisation, teaching effectiveness or psychometric validation claimed. AI editorial review only; no source code, models or installation executed, no production/status/publication changes.","adoption_requirements":"Public documentation can be read without installing. A separate tool trial needs current dependency/data-rights/provider review and human question/source checking; no learner data or model calls used here.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/ai-ml-questions-santana","sources":[{"label":"CristobalSantana/ai-ml-questions primary-readme","url":"https://raw.githubusercontent.com/CristobalSantana/ai-ml-questions/ec9fb1ed90102914761cb8c3f7dff326e5e73cdf/README.md","kind":"code-repository","is_primary":1},{"label":"CristobalSantana/ai-ml-questions statistics","url":"https://raw.githubusercontent.com/CristobalSantana/ai-ml-questions/ec9fb1ed90102914761cb8c3f7dff326e5e73cdf/docs/statistics.md","kind":"project-documentation","is_primary":1}]},{"slug":"ai-native-biomolecular-engineering-laboratory","title":"AI-native laboratory for autonomous biomolecular engineering","summary":"A research laboratory architecture co-designs models, experiments and instruments for multi-user, multi-objective nucleic-acid experimentation.","resource_type":"framework","disciplines":["biology","biomedicine"],"tasks":["experiment-design","experiment-execution","optimization","workflow-orchestration"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Platform-level automation is author-described; workflow level left unspecified because scope and researcher responsibilities depend on the implementation. Human safety and research oversight remain required.","source_checked_date":"2026-10-17","limitations":"Abstract-only canonical source review, with a separate partially read paper extraction. Author claims of autonomy, expert-level results and improved efficiency are not independently verified. The paper excerpt mentions reagent replenishment, so autonomy must not be interpreted as removal of all human work or safety oversight. No clinical validation, public service availability, software release or licence has been verified.","adoption_requirements":"Editorial, untested guidance: requires a suitably equipped molecular-biology laboratory, validated instrument interfaces, reagent management and qualified biosafety oversight. Protect sensitive genetic/sample data and validate protocols and outputs before research use; this is not clinical-use guidance.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/ai-native-biomolecular-engineering-laboratory","sources":[{"label":"An AI-native experimental laboratory for autonomous biomolecular engineering","url":"https://arxiv.org/abs/2507.02379","kind":"arxiv","is_primary":1}]},{"slug":"ai-radar-research","title":"ai_radar: AI economics and engineering research radar","summary":"Repository-documented research-monitoring workflow using domain-specific Claude Code skills to discover, check, deduplicate and archive public-source AI economics and engineering literature as structured Markdown entries.","resource_type":"workflow-process","disciplines":["computer-science","economics","social-science"],"tasks":["knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Workflow level left unspecified for this resource-level listing. Researchers choose scope, inspect inputs and methods, and verify generated outputs before reuse.","source_checked_date":"2026-10-17","limitations":"README-only review; scheduled operation, safety guarantees, test totals and library size are author claims not independently verified here. Coverage reflects configured profiles and accessible public sources, not a comprehensive literature search. Automated source checks and scoring do not replace researcher appraisal.","adoption_requirements":"Documented setup expects Claude subscription access, a user-configured GitHub Actions secret, interest profiles and an exact-host egress allowlist. Review hosted-workflow permissions and third-party content rights before reuse; no credentials or schedules were configured in this review.","updated_at":"2026-10-17T14:30:44Z","url":"https://research.wusong.net/cases/ai-radar-research","sources":[{"label":"ai_radar: AI economics and engineering research radar — primary repository README","url":"https://github.com/harryaquijeballon/ai_radar","kind":"code-repository","is_primary":1}]},{"slug":"ai-researcher","title":"AI-Researcher: an autonomous system for end-to-end AI research with Scientist-Bench","summary":"The authors present AI-Researcher, a system described as orchestrating literature review, hypothesis generation, algorithm implementation and manuscript preparation with minimal human intervention, alongside Scientist-Bench.","resource_type":"framework","disciplines":["computer-science"],"tasks":["hypothesis-generation","literature-research","research-coding","scientific-writing"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"The authors describe a complete research pipeline; detailed iteration and human-oversight boundaries were not inspected, so workflow level remains unspecified.","source_checked_date":"2026-10-17","limitations":"Abstract-only evidence. Benchmark construction, baselines and the quality rubric were not inspected, and the reported success rates are author-reported rather than replicated.","adoption_requirements":"The abstract page links code, but hardware, model access, costs, privacy and setup requirements were not reviewed.","updated_at":"2026-10-17T14:17:23Z","url":"https://research.wusong.net/cases/ai-researcher","sources":[{"label":"https://arxiv.org/abs/2505.18705","url":"https://arxiv.org/abs/2505.18705","kind":"arxiv","is_primary":1}]},{"slug":"ai-scientist","title":"The AI Scientist: towards end-to-end automation of AI research","summary":"A pipeline that automates the entire scientific life cycle end to end: it creates research ideas, writes code, runs experiments, plots and analyses data, writes the scientific manuscript, and performs its own peer review. A manuscript it generated passed the first round of peer review at a workshop of a top-tier ML conference.","resource_type":"use-case","disciplines":["computer-science","multidisciplinary"],"tasks":["automated-review","data-analysis","experiment-design","experiment-execution","hypothesis-generation","scientific-writing"],"evidence_level":"peer-reviewed","outcome_verification":"externally-validated","workflow_level":"L4","workflow_note":"Iterative idea generation, code execution, experiments, analysis and manuscript evaluation within a research remit. Humans set objectives and compute limits, supervise execution and must review claims before dissemination; automated peer review is not human oversight.","source_checked_date":"2026-10-15","limitations":"Reported results are from ML-research tasks; the peer-review outcome is a single workshop first-round pass, not a full publication. The system's own internal review step is automated and does not substitute for human review. Not independently reproduced.","adoption_requirements":"Frontier model API access with meaningful spend; ML experiment infrastructure; a human to screen outputs before anything leaves the lab, given the authors' own warnings about review load and literature noise.","updated_at":"2026-10-15T17:25:09Z","url":"https://research.wusong.net/cases/ai-scientist","sources":[{"label":"Towards end-to-end automation of AI research, Nature (2026)","url":"https://doi.org/10.1038/s41586-026-10265-5","kind":"doi","is_primary":1},{"label":"Code repository (The AI Scientist)","url":"https://github.com/SakanaAI/AI-Scientist","kind":"code-repository","is_primary":0}]},{"slug":"ai-trends-public-series-reference","title":"AI Trends: public-series research reference dashboard","summary":"A documented dashboard aggregates public time-series relevant to AIresearch context and governance.","resource_type":"tool","disciplines":["computer-science","economics","social-science"],"tasks":["data-management","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level assigned. The dashboard is a cited public-series research reference for AI compute, economics, research and adoption. Deterministic collection/charting and daily-job branding do not establish AI assistance or researcher output validation in one workflow.","source_checked_date":"2026-10-19","limitations":"README only. Live charts, data, caches, registry/code, CI runs, upstream studies and linked energy measurements were not audited. Most numbers belong to upstream sources, not this project’s measurements; the energy series has a separate origin. Provider, licence, coverage, count and freshness claims are documentation only. Append-only series cannot backfill; historical series can revise, and geographic/sparse coverage limits comparison. No research benefit, investment, audience/growth or causal claim is established. Moving HEAD is unpinned and non-atomic.","adoption_requirements":"Untested suggestion: inspect each primary series, date, definition, coverage and reuse terms; retain collection mode, failures and staleness in any research use.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/ai-trends-public-series-reference","sources":[{"label":"Project README snapshot","url":"https://raw.githubusercontent.com/stevenfazzio/ai-trends/HEAD/README.md","kind":"code-repository","is_primary":1}]},{"slug":"ai-wei-writing-study","title":"AI flavor in writing: literature synthesis and bounded prose comparison","summary":"A Chinese/English writing-research repository documents an AI-assisted literature synthesis, diagnostic tool and bounded Chinese-prose comparison with explicit methodological limits.","resource_type":"use-case","disciplines":["computer-science","humanities","linguistics"],"tasks":["data-analysis","knowledge-synthesis","scientific-writing"],"evidence_level":"repository-documented","outcome_verification":"author-reported","workflow_level":"L2","workflow_note":"Scoped L2 for author-defined prose sample/topics, two prompt conditions, model generation and diagnostic/comparative analysis. Literature/tool/experiment assistance and source checking were AI-performed; no peer review or human blind test. Limited author-reported results are not an authorship detector, general writing-quality measure or independent reproduction.","source_checked_date":"2026-10-19","limitations":"Full README and selected experiment chapter read as timestamped moving-main-branch documents; commit lookup hit API quota, so no commit pin/atomic version guarantee. No original studies, code, model outputs, corpus, model/Bedrock calls or statistics reproduced. Source reports13human pieces and78model texts; sample is small/special, truncated human versus complete model text, one genre/language and no human blind test. Qualitative comparisons/AI source checking are not expert validation. Source-stated model names and rights/legal statements not independently verified or offered as current facts/advice. No detector-evasion or concealment guidance; source frames improvement/disclosure and cautions against score chasing. AI editorial review only; no source code, models or installation executed, no production/status/publication changes.","adoption_requirements":"Read the limitations before any trial; diagnostic tool is described as Python3.9+ standard-library based, while reproducing the experiment has separate dependencies/provider permissions. No installation, corpus download or model call performed. Review rights/disclosure independently.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/ai-wei-writing-study","sources":[{"label":"tsaol/ai-wei primary-readme","url":"https://raw.githubusercontent.com/tsaol/ai-wei/main/README.md","kind":"code-repository","is_primary":1},{"label":"tsaol/ai-wei experiment-methods","url":"https://raw.githubusercontent.com/tsaol/ai-wei/main/docs/06-experiment.md","kind":"project-documentation","is_primary":1}]},{"slug":"ai4sarcopenia-literature-daily","title":"AI4Sarcopenia Literature Daily","summary":"Documented automated literature-monitoring workflow for AI-driven body shape analysis for sarcopenia, with daily updates and a linked website.","resource_type":"workflow-process","disciplines":["biomedicine","medicine"],"tasks":["literature-research"],"evidence_level":"repository-documented","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Keeps a domain researcher current: daily retrieval, screening and summarisation of newly indexed literature.","source_checked_date":"2026-10-17","limitations":"Digests are machine-generated without expert appraisal; screening/summary quality author-described; not executed during this review.","adoption_requirements":"Source available under MIT; domain adaptation requires configuration. Full runtime/provider requirements are not established by the saved excerpts.","updated_at":"2026-10-17T09:28:23Z","url":"https://research.wusong.net/cases/ai4sarcopenia-literature-daily","sources":[{"label":"AI4SarcopeniaLiteratureDaily - GitHub repository (Aizierjiang)","url":"https://github.com/Aizierjiang/AI4SarcopeniaLiteratureDaily","kind":"code-repository","is_primary":1}]},{"slug":"alabos","title":"AlabOS: reconfigurable autonomous-laboratory orchestration","summary":"AlabOS is a Python-based framework for orchestrating experiments and managing resources in automated materials synthesis and characterization laboratories.","resource_type":"framework","disciplines":["materials-science"],"tasks":["experiment-execution","workflow-orchestration"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Resource-level workflow classification unspecified; an instantiated workflow depends on the research task, implementation and oversight. Researchers retain responsibility for validation and safe use.","source_checked_date":"2026-10-17","limitations":"Abstract-only review of arXiv v2; full methods, software distribution, installation and conflict-avoidance behavior were not inspected or tested. The arXiv record links a journal article (10.1039/D4DD00129J), whose publisher body was not read. Deployment scale is author-reported, not independently reproduced. Orchestration is useful infrastructure for AI-assisted research, not evidence of an AI model inside the framework.","adoption_requirements":"Untested editorial guidance: adaptation would require laboratory automation expertise, equipment interfaces and researcher-approved experimental/safety constraints; exact dependencies and hardware compatibility are unverified.","updated_at":"2026-10-17T15:13:29Z","url":"https://research.wusong.net/cases/alabos","sources":[{"label":"AlabOS: A Python-based Reconfigurable Workflow Management Framework for Autonomous Laboratories — primary arXiv record","url":"https://arxiv.org/abs/2405.13930","kind":"arxiv","is_primary":1}]},{"slug":"alphaevolve","title":"AlphaEvolve","summary":"AlphaEvolve evolves programs through repeated proposals and executable evaluation. The paper reports a 48-scalar-multiplication construction for two 4×4 complex matrices, compared with 49 for recursive Strassen. The public repository contains mathematical results and verifier code, explicitly excluding the agent itself. These bounded constructions do not establish general scientific autonomy or independently replicated system performance.","resource_type":"framework","disciplines":["mathematics"],"tasks":["optimization","research-coding"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":"L3","workflow_note":"L3 applies only to the user-specified program-search workflow. Users provide initial code, evaluator code and optional configuration; authors inspect the resulting mathematical constructions. Automated scores constrain proposals; executable evaluation is required. The full agent implementation is not in the public result repository. Stops and safety are evaluator/configuration dependent; this record does not certify every possible evolved program or broader autonomous science.","source_checked_date":"2026-10-19","limitations":"Tasks require programmable evaluators. Public results/verifier notebook explicitly excludes the AlphaEvolve system code. Mathematical artifact verification is distinct from independent replication of agent performance; no code was executed here.","adoption_requirements":"Untested suggestion: inspect the pinned sources and validate data, evaluator assumptions, implementation availability and researcher controls before planning a bounded trial. No setup or trial was executed.","updated_at":"2026-10-20T08:01:56Z","url":"https://research.wusong.net/cases/alphaevolve","sources":[{"label":"Primary manuscript — 2506.13131v1","url":"https://arxiv.org/html/2506.13131v1","kind":"arxiv","is_primary":1},{"label":"Primary project repository or artifact (version limits apply)","url":"https://github.com/google-deepmind/alphaevolve_results/blob/4226acbf237ff9ad10ba7673a2af127a2d8a5971/README.md","kind":"code-repository","is_primary":1}]},{"slug":"alphaflow-reinforcement-learning-fluidic-lab","title":"AlphaFlow: reinforcement-learning-guided multi-step synthesis","summary":"AlphaFlow combines reinforcement learning with a modular microdroplet reactor and in-situ spectral monitoring to discover and optimize multi-step core-shell semiconductor nanoparticle synthesis. The authors report a route outperforming conventional sequences; this review did not reproduce that result.","resource_type":"use-case","disciplines":["chemistry","materials-science"],"tasks":["experiment-design","experiment-execution","optimization"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":"L3","workflow_note":"Bounded closed-loop experiment selection and execution within a researcher-defined apparatus and chemical task; researchers supply objectives, equipment, initial materials, safety limits and final interpretation. L3 does not mean unrestricted or unsupervised research autonomy.","source_checked_date":"2026-10-17","limitations":"Primary abstract-level review with secondary news used only to identify the resource. Full HTML was captured but methods, supplementary data and quantitative outcome claims were not comprehensively adjudicated. No reproduction or software execution. Hardware, chemical handling and nanoparticle waste require specialist assessment; repository identity is not verification of code or licence.","adoption_requirements":"Specialist chemical laboratory, compatible automation and measurement hardware, trained operators, local risk assessment and institution-approved operating limits; evaluate equipment/software rights and data confidentiality before reuse. Costs, licences and setup completeness were not audited.","updated_at":"2026-10-17T19:24:37Z","url":"https://research.wusong.net/cases/alphaflow-reinforcement-learning-fluidic-lab","sources":[{"label":"AlphaFlow: autonomous discovery and optimization of multi-step chemistry using a self-driven fluidic lab guided by reinforcement learning","url":"https://www.nature.com/articles/s41467-023-37139-y","kind":"doi","is_primary":1},{"label":"This self-driving lab learns to make nanoparticles all on its own","url":"https://doi.org/10.47287/cen-10113-leadcon","kind":"doi","is_primary":0}]},{"slug":"alphageometry","title":"AlphaGeometry: solving olympiad geometry without human demonstrations","summary":"A neuro-symbolic theorem prover for Euclidean plane geometry that sidesteps the shortage of human training data by synthesizing millions of theorems and proofs. On a test set of 30 latest olympiad-level problems the authors report it solves 25, versus 10 for the previous best method.","resource_type":"use-case","disciplines":["mathematics"],"tasks":["proof-search"],"evidence_level":"peer-reviewed","outcome_verification":"machine-verified","workflow_level":"L3","workflow_note":"Neural proposals guide a symbolic deduction tool through multi-step proof search. Humans supply formalized problems and assess applicability; this is bounded theorem proving, not an autonomous research programme.","source_checked_date":"2026-10-15","limitations":"Scope is Euclidean plane geometry; problems must be encoded into the system's formal language. The 25/30 result is an author-reported evaluation on one curated test set. Not independently reproduced.","adoption_requirements":"Familiarity with formal geometry languages and the system's input format; compute for running the neural guidance model; suitable mainly for research groups studying automated reasoning rather than general mathematics workflows.","updated_at":"2026-10-15T17:25:09Z","url":"https://research.wusong.net/cases/alphageometry","sources":[{"label":"Trinh et al., Solving olympiad geometry without human demonstrations, Nature (2024)","url":"https://doi.org/10.1038/s41586-023-06747-5","kind":"doi","is_primary":1},{"label":"Code repository (AlphaGeometry)","url":"https://github.com/google-deepmind/alphageometry","kind":"code-repository","is_primary":0}]},{"slug":"alphakit-investment-research-skills","title":"AlphaKit: bilingual investment-research skills","summary":"AlphaKit documents a bilingual finance and investment-research skill collection. Its original skills cover incremental business sizing, model exhibits and preliminary stock-research briefs; the sizing workflow accepts sourced assumptions and produces scenarios and sensitivity outputs.","resource_type":"agent-skill","disciplines":["business","economics"],"tasks":["data-analysis","knowledge-synthesis"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation is unspecified. Researchers choose goals, inputs, permissions and evaluation criteria and remain responsible for interpreting outputs.","source_checked_date":"2026-10-17","limitations":"README and original-skills documentation only; no skill execution, investment outcome verification, host integration test or underlying third-party catalogue audit. The source explicitly distinguishes source checks from execution checks; examples are synthetic. Documentation reports other hosts' discovery/calling has not been tested. No return or predictive-accuracy claim is accepted.","adoption_requirements":"A compatible custom-skill host and complete skill folders; Python for sizing, Pillow/openpyxl and Chinese-capable fonts for rendering, as documented. Use only authorized source material, preserve package and third-party licences, and review permissions and assumptions before running.","updated_at":"2026-10-17T20:39:39Z","url":"https://research.wusong.net/cases/alphakit-investment-research-skills","sources":[{"label":"AlphaKit: bilingual investment-research skills — README","url":"https://github.com/Jeromelll/AlphaKit","kind":"code-repository","is_primary":1},{"label":"AlphaKit: bilingual investment-research skills — original skills documentation","url":"https://github.com/Jeromelll/AlphaKit/blob/main/docs/RESEARCH_SKILLS.md","kind":"code-repository","is_primary":1}]},{"slug":"alphaproof","title":"AlphaProof: olympiad-level formal mathematical reasoning with reinforcement learning","summary":"An AlphaZero-inspired reinforcement-learning agent that learns to find formal proofs in Lean by training on millions of auto-formalized problems. At IMO 2024, as the core reasoning engine of Google DeepMind's system, it solved three of the five non-geometry problems; combined with AlphaGeometry 2 this reached a silver-medal-equivalent score.","resource_type":"use-case","disciplines":["mathematics"],"tasks":["proof-search"],"evidence_level":"peer-reviewed","outcome_verification":"machine-verified","workflow_level":"L3","workflow_note":"Multi-step proof search uses Lean feedback and problem-specific adaptation. Humans supply formalized objectives and evaluate relevance. Iteration inside a bounded proof task is not, alone, end-to-end autonomous research.","source_checked_date":"2026-10-15","limitations":"Requires multi-day compute for competition-level problems and a Lean formalization step for the problem statement. Reported competition performance is an author evaluation; broader mathematical utility beyond olympiad-style problems is not established. Not independently reproduced.","adoption_requirements":"Lean 4 environment and auto-formalization tooling; very large compute budgets for training and inference; skills at the intersection of proof engineering and reinforcement learning. Practical today mostly for groups studying automated reasoning.","updated_at":"2026-10-15T17:25:09Z","url":"https://research.wusong.net/cases/alphaproof","sources":[{"label":"Hubert et al., Olympiad-level formal mathematical reasoning with reinforcement learning, Nature (2026; online 2025-11-12)","url":"https://doi.org/10.1038/s41586-025-09833-y","kind":"doi","is_primary":1}]},{"slug":"antevorta-ai-context-project-guide","title":"AI Context Project Guide","summary":"Advisory guide for organizing private context repositories, workspace rules and initiative tracking across AI-assisted work sessions, including research projects.","resource_type":"tutorial-template","disciplines":["multidisciplinary"],"tasks":["data-management","workflow-orchestration"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level workflow level unspecified; documented configuration is not verified autonomous operation. Humans control scope and inspect outputs.","source_checked_date":"2026-10-17","limitations":"README-only review; linked primer, advisor specification, patterns and skills were not inspected or executed. No evaluation of productivity or context recovery. README discloses deferred Claude Code companion documents and known practice/documentation drift.","adoption_requirements":"An AI-enabled editor or assistant and a private context repository. Check provider data retention, intellectual-property and research-data policies before loading material; retain researcher approval over changes and tool actions.","updated_at":"2026-10-17T17:46:50Z","url":"https://research.wusong.net/cases/antevorta-ai-context-project-guide","sources":[{"label":"AI Context Project Guide — primary README","url":"https://gitlab.com/antevorta/ai-context-project-guide/-/raw/main/README.md","kind":"code-repository","is_primary":1}]},{"slug":"archibate-agent-skills-cpp-hpc","title":"archibate agent-skills: C++ and HPC coding guidance","summary":"A repository of agent skills whose README documents C++ coding guidance and a cpp-hpc-optimization skill covering profiling, parallelism, SIMD, sparse matrices and memory-aware optimisation. It prescribes performance comparisons and correctness checks before and after changes.","resource_type":"agent-skill","disciplines":["computer-science","engineering"],"tasks":["optimization","research-coding"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation is unspecified. Researchers define the task, inspect sources/code and validate outputs; this review did not execute a workflow.","source_checked_date":"2026-10-17","limitations":"README-only substantive review, supplemented by repository identity metadata. SKILL.md contents, installer, correctness tests, benchmarks, dependencies and security claims were not audited or executed. Advertised optimisation practices are instructions, not measured improvements.","adoption_requirements":"Requires a compatible agent environment and an appropriate C++ build/profiling setup. README names Codex, OpenCode, Claude Code and Pi. Inspect skill and installer contents before adoption; do not execute a remote shell installer blindly or expose research secrets.","updated_at":"2026-10-17T17:46:50Z","url":"https://research.wusong.net/cases/archibate-agent-skills-cpp-hpc","sources":[{"label":"archibate agent-skills: C++ and HPC coding guidance — primary documentation","url":"https://raw.githubusercontent.com/archibate/agent-skills/master/README.md","kind":"code-repository","is_primary":1}]},{"slug":"aria-causal-aware-materials-reasoning","title":"ARIA: causal-aware reasoning for materials discovery","summary":"ARIA is a causal-aware framework that conditions language-model use of materials knowledge on the completeness of process-structure-property evidence.","resource_type":"framework","disciplines":["materials-science"],"tasks":["hypothesis-generation","knowledge-synthesis","materials-discovery"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Workflow level left unspecified: the abstract does not establish operational boundaries and researcher responsibilities sufficiently for a level claim. Researchers remain responsible for domain validation, safety and interpretation.","source_checked_date":"2026-10-17","limitations":"Abstract-only primary-source review. No software, hardware, dataset or reported outcome independently reproduced; detailed methods, evaluation settings, distribution and licensing were not inspected. The arXiv record links an ACM DOI, but the publisher article and peer-review status were not independently checked.","adoption_requirements":"Editorial, untested guidance: inspect knowledge-graph extraction, model and literature-search requirements in the paper; preserve evidence provenance, respect article rights and review causal chains and generated designs with materials expertise. No public knowledge-graph download or runnable implementation was verified.","updated_at":"2026-10-17T15:13:29Z","url":"https://research.wusong.net/cases/aria-causal-aware-materials-reasoning","sources":[{"label":"ARIA: A Causal-Aware Framework for Rescuing LLM Reasoning in Trustworthy Materials Discovery","url":"https://arxiv.org/abs/2606.22375","kind":"arxiv","is_primary":1}]},{"slug":"ario-auditable-epistemic-framework","title":"Ario — Auditable Epistemic Framework for AI","summary":"Experimental, repository-documented framework for investigating AI identity, memory and revision by separating observations, evidence, inferences and claims, and preserving the history of changed assessments.","resource_type":"framework","disciplines":["computer-science","humanities"],"tasks":["agent-evaluation","data-management"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation unspecified. Researchers approve objectives and inspect outputs; documentation is not an executed workflow.","source_checked_date":"2026-10-17","limitations":"README-only source review by an AI editorial agent; no implementation execution, independent outcome evaluation, security audit or full linked-document review. HEAD is mutable; saved content and hashes preserve the inspected version. Official site and GitHub Pages mirror are explicit identity aliases. Zenodo 23259916 is linked as the preliminary Ario report. Record 23376521 and project-plus-book 23375401 are related identity candidates, not independently inspected deposits. Being With the Other book editions (23326995, 23375350, 23376577) are related works, not automatically identical to the software framework; ORCID identifies a person, not this resource.","adoption_requirements":"Researcher-defined evaluation questions, protected interaction records and review of implementation contracts before any execution. Python implementation and tests are mentioned in the README but were not inspected or run.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/ario-auditable-epistemic-framework","sources":[{"label":"SepehrGhanbari/Ario — README","url":"https://raw.githubusercontent.com/SepehrGhanbari/Ario/HEAD/README.md","kind":"code-repository","is_primary":1}]},{"slug":"aroughidea-autoresearch-starter","title":"autoresearch-starter: recorded small-model experiment loops","summary":"Template documenting small language-model experiments in which a coding agent edits a training recipe, runs time-bounded training and keeps or undoes changes while recording the session.","resource_type":"tutorial-template","disciplines":["computer-science","education"],"tasks":["experiment-execution","optimization","workflow-orchestration"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation left unspecified. Researchers/educators choose materials, govern model access and verify interpretations and outputs.","source_checked_date":"2026-10-17","limitations":"Complete current README reviewed only; no code, session records, demo or hardware instructions executed/audited. Queue metadata says five-minute experiments; the retrieved README defaults to ten minutes. Scores depend on hardware speed and the greedy loop may retain improvements smaller than baseline noise. Unattended coding-agent permissions can execute arbitrary user-level commands; use isolation, spending/runtime limits and researcher inspection. No demo gains, training performance, cost or overnight reliability verified.","adoption_requirements":"The documented Windows/NVIDIA quickstart needs git, uv, Python 3.10+ and an appropriate GPU: Turing 8 GB+ or newer listed generations 10 GB+. Agent-driven sessions additionally need a coding agent/account and Node.js. Review platform-specific hardware guidance and permissions before adoption.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/aroughidea-autoresearch-starter","sources":[{"label":"aroughidea/autoresearch-starter — primary README","url":"https://github.com/aroughidea/autoresearch-starter","kind":"code-repository","is_primary":1}]},{"slug":"artificial-whole-lab-orchestration","title":"Artificial: whole-lab orchestration and scheduling","summary":"Artificial is described as a whole-lab orchestration and scheduling system integrating AI-driven decision-making with coordination of instruments, robots and personnel for drug-discovery workflows.","resource_type":"framework","disciplines":["biomedicine","chemistry"],"tasks":["experiment-execution","workflow-orchestration"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level workflow automation intentionally unspecified. A reusable component/process reference is not a measured autonomous workflow; researcher oversight remains necessary.","source_checked_date":"2026-10-17","limitations":"Abstract-only substantive review. No runtime, interfaces, released code, deployment instructions, quantitative acceleration, reproducibility measurement or experimental validation inspected. No clinical efficacy or discovered therapeutic is established. The HTML search result mentions SLAS Technology, but no exact published record was confirmed by exact-title/byline searches and a saved Crossref title query; a manuscript venue line alone is not publication evidence. Preprint classification is provisional to the inspected record, not a claim that no later publication exists.","adoption_requirements":"Editorial, untested: obtain current vendor documentation and determine instrument interfaces, scheduling constraints, model access and laboratory safety requirements before evaluating fit. Availability, licence, pricing, BioNeMo access and compatibility were not audited.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/artificial-whole-lab-orchestration","sources":[{"label":"Accelerating drug discovery with Artificial: a whole-lab orchestration and scheduling system for self-driving labs","url":"https://arxiv.org/abs/2504.00986","kind":"arxiv","is_primary":1}]},{"slug":"asd-ml-methods-review-reference","title":"ASD machine-learning research: method-family review reference","summary":"A journal-identified review discusses families of hybrid ML methods and research-selection criteria in ASD research.","resource_type":"framework","disciplines":["computer-science","psychology"],"tasks":["experiment-design","literature-research"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level assigned. The reference maps ensembles, feature combinations and stacking in ASD-related ML research. The surveyed methods do not establish one connected researcher-supervised workflow or output-validation protocol.","source_checked_date":"2026-10-19","limitations":"Selected primary PDF passages and official arXiv record only. The full review, databases, data, clinical studies and results were not audited. PDF/arXiv metadata identify a Heliyon article; independent publisher and review-process verification was unavailable after a Crossref rate limit. The DOI year is not treated as the publication date. Cached parsing/raw-version parity remains unverified. No diagnostic or treatment efficacy, clinical adoption or privacy assurance is established.","adoption_requirements":"Untested suggestion: inspect a chosen method’s primary evidence, data, validation and clinical limitations before research reuse.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/asd-ml-methods-review-reference","sources":[{"label":"Selected journal-article passages","url":"https://arxiv.org/pdf/2608.18188","kind":"arxiv","is_primary":1}]},{"slug":"asreview","title":"ASReview: active-learning-assisted systematic review screening","summary":"An open-source machine-learning pipeline helps researchers screen titles and abstracts for systematic reviews and meta-analyses, with simulation and user-experience evaluations.","resource_type":"use-case","disciplines":["multidisciplinary"],"tasks":["knowledge-synthesis","literature-research"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":"L2","workflow_note":"Human-led screening workflow: researchers set eligibility criteria and label records; active learning reprioritizes the remaining literature. Humans retain inclusion and stopping decisions. Iterative machine learning here is not an autonomous research agent.","source_checked_date":"2026-10-15","limitations":"Simulation and user-experience findings do not guarantee recall or time savings for a new review. Human eligibility judgments and stopping decisions remain necessary.","adoption_requirements":"A literature record collection, explicit eligibility criteria, reviewer time for labeling and a defensible stopping and audit procedure.","updated_at":"2026-10-15T17:25:09Z","url":"https://research.wusong.net/cases/asreview","sources":[{"label":"van de Schoot et al., An open source machine learning framework for efficient and transparent systematic reviews, Nature Machine Intelligence (2021)","url":"https://www.nature.com/articles/s42256-020-00287-7","kind":"doi","is_primary":1},{"label":"ASReview primary repository","url":"https://github.com/asreview/asreview","kind":"code-repository","is_primary":1}]},{"slug":"atlasti-ai-features","title":"ATLAS.ti AI features","summary":"ATLAS.ti describes AI-assisted document conversations, intention-guided coding and audio/video transcription for qualitative research.","resource_type":"tool","disciplines":["business","humanities","social-science"],"tasks":["data-analysis","knowledge-synthesis","transcription"],"evidence_level":"vendor-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Vendor homepage, not a comparative evaluation. No claimed productivity percentages are adopted. Interpretive validity, consent, data processing location and transcription quality require user checks.","adoption_requirements":"Applicable ATLAS.ti service/features, consented research materials and a human qualitative-analysis protocol.","updated_at":"2026-10-16T14:58:30Z","url":"https://research.wusong.net/cases/atlasti-ai-features","sources":[{"label":"ATLAS.ti | The #1 Software for Qualitative Data Analysis - ATLAS.ti","url":"https://atlasti.com","kind":"vendor-documentation","is_primary":1}]},{"slug":"auditable-economics-loop","title":"An auditable AI agent loop for empirical economics","summary":"A working-paper workflow combining logged agent searches with post-search holdout evaluation for adaptive empirical analysis.","resource_type":"workflow-process","disciplines":["business","economics"],"tasks":["data-analysis","research-coding","workflow-orchestration"],"evidence_level":"institutional-report","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level entry: no universal autonomy level is assigned. The researcher defines the task, controls access and validates outputs; classify an actual instantiated workflow separately.","source_checked_date":"2026-10-15","limitations":"Only the official working-paper summary was reviewed, not the full paper, code or experiments. The summary reports benchmark improvements by the authors; no magnitude, independent reproduction or general research-efficiency benefit is established here.","adoption_requirements":"A research question, reproducible data/code, logged searches and a genuine untouched evaluation set. Full paper and implementation need review before adoption.","updated_at":"2026-10-15T17:25:09Z","url":"https://research.wusong.net/cases/auditable-economics-loop","sources":[{"label":"Reviewed primary source: An auditable AI agent loop for empirical economics","url":"https://www.philadelphiafed.org/the-economy/an-auditable-ai-agent-loop-for-empirical-economics-a-case-study-in-forecast-combination","kind":"institutional-report","is_primary":1}]},{"slug":"audited-ai-archival-research-acid-attack-violence","title":"Audited AI-assisted archival research on acid-attack violence","summary":"A methodological case study describes AI assistants supplementing archival searches on acid-attack violence, with standardized model and prompt choices, session logging and multi-stage auditing. The author reports descriptive analysis of 417 sessions across 212 WHO-reporting country cases.","resource_type":"workflow-process","disciplines":["social-science"],"tasks":["archival-research","data-extraction","knowledge-synthesis"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Researcher-led workflow: people set research questions, approve data access and adjudicate outputs; no autonomous end-to-end performance claim. Workflow level intentionally unspecified.","source_checked_date":"2026-10-17","limitations":"Abstract-only publisher review; extraction returned an incomplete introduction, which is not used to infer methods. No exact model versions, prompts, audit error rates, dataset access or executable recipe verified. Retrospective descriptive case study, not independent replication. Sensitive violence-related records require rights, privacy and contextual checks.","adoption_requirements":"Access rights to archival sources, model access, session/provenance logging and domain-informed researchers able to audit multilingual evidence. Do not upload identifiable survivor information without an appropriate lawful and ethical basis.","updated_at":"2026-10-17T19:24:37Z","url":"https://research.wusong.net/cases/audited-ai-archival-research-acid-attack-violence","sources":[{"label":"Utilizing AI Assistants in Social Science Research: A Methodological Case Study of AI-Assisted Archival Research on Acid-Attack Violence","url":"https://www.mdpi.com/2076-0760/15/9/616","kind":"doi","is_primary":1}]},{"slug":"autism-emotion-sensor-ml-review-reference","title":"Emotion-recognition research in autism: sensor and ML review reference","summary":"A journal review organises research questions around sensing technologies, ML techniques and limitations in autism emotion-recognition research.","resource_type":"framework","disciplines":["computer-science","psychology"],"tasks":["experiment-design","literature-research"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level assigned. The review provides a sensor/ML method-selection reference; its checked search and screening procedure is human performed. It does not establish an AI-assisted review or one supervised AI research workflow.","source_checked_date":"2026-10-19","limitations":"Selected PDF methods/excerpts and official abstract only. The full review, tables, primary studies, data and medical outcomes were not audited. DOI metadata supports article identity, not independent result validation. Cached parsing and raw PDF/version parity remain unverified. No clinical, diagnostic, treatment, privacy or emotion-inference validity recommendation is established.","adoption_requirements":"Untested suggestion: use the review to locate a specific primary method, then examine its evidence and limitations independently before research reuse.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/autism-emotion-sensor-ml-review-reference","sources":[{"label":"Selected review methods and primary paper passages","url":"https://arxiv.org/pdf/2407.04712","kind":"arxiv","is_primary":1}]},{"slug":"auto-empirical-research-skills-aers","title":"Auto-Empirical Research Skills (AERS)","summary":"Repository of reusable agent skills and research-workflow guidance for social-science empirical work, including literature review, data preparation, identification strategies, estimation and robustness checks.","resource_type":"agent-skill","disciplines":["economics","social-science"],"tasks":["data-analysis","literature-research","scientific-writing"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Workflow level left unspecified for this resource-level listing. Researchers choose scope, inspect inputs and methods, and verify generated outputs before reuse.","source_checked_date":"2026-10-17","limitations":"Partial README review, not a skill-by-skill, benchmark or security audit. Queue marketing counts differ from the current README and are not repeated as verified inventory. Claimed institutional affiliations, automated paper quality and time savings are unverified. Some advertised components aim to evade AI detection or remove provenance; they are not recommended. Preserve AI-use disclosure, source attribution and research integrity.","adoption_requirements":"A compatible agent environment and task-specific statistical software/data access; inspect each selected skill, upstream licence, executable content and tool permissions before installation. Use project-local scope and protect sensitive research data.","updated_at":"2026-10-17T14:30:44Z","url":"https://research.wusong.net/cases/auto-empirical-research-skills-aers","sources":[{"label":"Auto-Empirical Research Skills (AERS) — primary repository README","url":"https://github.com/brycewang-stanford/Auto-Empirical-Research-Skills","kind":"code-repository","is_primary":1}]},{"slug":"auto-ta","title":"Auto-TA: Towards Scalable Automated Thematic Analysis (TA) via Multi-Agent Large Language Models with Reinforcement Learning","summary":"Automated multi-agent LLM pipeline performing end-to-end thematic analysis on clinical narratives, applied to congenital heart disease narratives, with optional RLHF.[19]","resource_type":"framework","disciplines":["biomedicine","medicine"],"tasks":["data-analysis","research-coding"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-16","limitations":"Fully automated design removes manual coding; scalability and theme-quality claims are author-reported. The RLHF component is optional. Abstract-level review only - full-text methods and any released materials were not inspected, and no independent performance validation.","adoption_requirements":"Multi-agent LLM pipeline setup; clinical narrative data governance; optional RLHF components.","updated_at":"2026-10-17T06:33:49Z","url":"https://research.wusong.net/cases/auto-ta","sources":[{"label":"Auto-TA: Towards Scalable Automated Thematic Analysis (TA) via Multi-Agent Large Language Models with Reinforcement Learning","url":"https://arxiv.org/abs/2506.23998","kind":"arxiv","is_primary":1}]},{"slug":"autoaurora-student-code-completion-study","title":"AutoAurora: student perspectives on AI code completion","summary":"An educational research reference based on ten student interviews about AI code completion, using the authors’ StarCoder-powered AutoAurora Visual Studio Code extension as a research instrument.","resource_type":"tutorial-template","disciplines":["computer-science","education"],"tasks":["research-coding"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Resource-level automation classification unspecified. This is a research/reference listing, not an exercised workflow; researchers select objectives and check interpretation and outputs.","source_checked_date":"2026-10-17","limitations":"Abstract-only substantive review, plus institutional publication metadata and official workshop review policy. This is a qualitative student-perspective study, not a causal estimate of learning gains or measured productivity improvement. No extension installation, code audit, pedagogical trial or independent replication was performed. Detailed interview procedures, recruitment and analytic checks were not inspected. The abstract calls the extension open-source; its actual license and canonical distribution were not audited.","adoption_requirements":"Editorial, untested use: read the study as a teaching/research-design reference. Any new student study needs a suitable coding environment, consent and privacy safeguards, and an independently checked instrument; installation requirements are not established here.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/autoaurora-student-code-completion-study","sources":[{"label":"AutoAurora student-perspective study — arXiv abstract","url":"https://arxiv.org/abs/2311.00177","kind":"arxiv","is_primary":1},{"label":"Monash University — official conference publication record","url":"https://research.monash.edu/en/publications/students-perspectives-on-ai-code-completion-benefits-and-challeng/","kind":"institutional-report","is_primary":1},{"label":"COMPSAC 2024 — workshop review policy","url":"https://ieeecompsac.computer.org/2024/call-for-workshop-proposals/","kind":"workshop-page","is_primary":1}]},{"slug":"autogen","title":"AutoGen","summary":"Multi-agent application framework usable to prototype research assistants; now in maintenance mode per its README.[16]","resource_type":"framework","disciplines":["computer-science","multidisciplinary"],"tasks":["workflow-orchestration"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Generic framework, not demonstrated research gains. Maintenance mode: new work should consider migration to the successor; historical compatibility is not a current recommendation. Documentation review only - no software execution or independent performance validation.","adoption_requirements":"Programming environment, model/tool integration and sandboxing; consider migration for new projects.","updated_at":"2026-10-17T06:33:49Z","url":"https://research.wusong.net/cases/autogen","sources":[{"label":"AutoGen - repository (microsoft/autogen)","url":"https://github.com/microsoft/autogen","kind":"code-repository","is_primary":1}]},{"slug":"autonomous-chemical-laboratories-comparison-reference","title":"Autonomous chemical laboratories: architecture comparison reference","summary":"A review offers an architecture comparison for AI-assisted chemical laboratories.","resource_type":"framework","disciplines":["chemistry","materials-science","robotics"],"tasks":["experiment-design","knowledge-synthesis","workflow-orchestration"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level assigned. The review compares hardware, software and integration choices across laboratories. Its different systems do not establish one common human-control, stopping or conclusion-validation protocol. The comparison is a secondary reference, not a common benchmark.","source_checked_date":"2026-10-19","limitations":"Selected PDF text and publisher identity only; raw PDF/version parity, figures, full references and experiments were not audited. Comparison ratings and numerical outcomes were not validated. A heading/group mismatch, citation-number inconsistencies and untraced performance claims limit attribution. The journal states single-blind review; individual review conduct and quality were not audited. No result or implementation is inherited from the surveyed laboratories.","adoption_requirements":"Untested suggestion: verify each selected underlying study, equipment interface, control, cost and reuse term before any trial.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/autonomous-chemical-laboratories-comparison-reference","sources":[{"label":"Primary review abstract and identity","url":"https://doi.org/10.4208/cicc.2025.90.03","kind":"doi","is_primary":1},{"label":"Selected publisher PDF architecture and limitations","url":"https://www.global-sci.com/api/galley-access/35409/download?returnTo=%2Fcicc%2Farticle%2Fview%2F22828%23full-text","kind":"doi","is_primary":1}]},{"slug":"autonomous-chemistry-closed-loop-reference-highlight","title":"Autonomous chemistry: closed-loop components and limitations reference","summary":"A Research Highlight maps planning, robotic synthesis, analytical feedback and AI-assisted route refinement, with limitations for designing autonomous chemistry workflows.","resource_type":"framework","disciplines":["chemistry","materials-science","robotics"],"tasks":["experiment-design","knowledge-synthesis","workflow-orchestration"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level assigned. The Research Highlight maps literature-informed planning, robotic synthesis, analytical feedback and route refinement, with data, transferability, hardware and recovery gaps. Its ideal cycle and heterogeneous examples do not establish one bounded remit with human goals, constraints, stopping rules and conclusion validation.","source_checked_date":"2026-10-19","limitations":"Selected publisher HTML main text, identity, declarations and limitations only. Figures, full cited studies, experiments, code and implementation were not audited. This is a Research Highlight/secondary synthesis, not a new laboratory. No yield, speed, generality, reliability or safety benefit is inherited from surveyed systems. The journal-source label relies on publication identity and stated policy, not audited review conduct, indexing or article quality. No chemistry experiment was performed.","adoption_requirements":"Untested reference suggestion: trace each selected component to its underlying primary study and specify human boundaries, chemical-safety review, failure recovery and result validation before a trial.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/autonomous-chemistry-closed-loop-reference-highlight","sources":[{"label":"Primary Research Highlight: selected main text and limitations","url":"https://doi.org/10.20517/cs.2025.66","kind":"doi","is_primary":1}]},{"slug":"autonomous-lab-orchestrator-scheduling","title":"Autonomous lab orchestrator scheduling method","summary":"Two-step method for resource utilization in autonomous laboratories: constraint-programming scheduling under hardware constraints plus status-dependency execution, demonstrated on an autonomous metal-organic-framework synthesis platform.[11]","resource_type":"framework","disciplines":["chemistry","materials-science"],"tasks":["experiment-execution","optimization","workflow-orchestration"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans set objectives and validate results.","source_checked_date":"2026-10-16","limitations":"Author-reported demonstration; not independently reproduced. Implementation availability not verified. Published version checked at record level (title/venue/affiliations).","adoption_requirements":"Integration into an autonomous-lab orchestration stack; constraint solver; hardware capacity/throughput data.","updated_at":"2026-10-17T08:00:32Z","url":"https://research.wusong.net/cases/autonomous-lab-orchestrator-scheduling","sources":[{"label":"Optimal Resource Utilization for Autonomous Laboratory Orchestrators - arXiv:2607.01188","url":"https://arxiv.org/abs/2607.01188","kind":"arxiv","is_primary":1}]},{"slug":"autonomous-scanning-droplet-cell-alloy-design","title":"Autonomous scanning droplet cell for alloy-coating design","summary":"An autonomous scanning droplet cell platform supports on-demand alloy electrodeposition and real-time electrochemical characterization of multicomponent alloys.","resource_type":"framework","disciplines":["materials-science"],"tasks":["experiment-execution","materials-discovery","optimization"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Workflow level left unspecified: the abstract does not establish operational boundaries and researcher responsibilities sufficiently for a level claim. Researchers remain responsible for domain validation, safety and interpretation.","source_checked_date":"2026-10-17","limitations":"Abstract-only primary-source review. No software, hardware, dataset or reported outcome independently reproduced; detailed methods, evaluation settings, distribution and licensing were not inspected.","adoption_requirements":"Editorial, untested guidance: assess electrodeposition and electrochemical-measurement equipment, alloy chemistry expertise, calibration and laboratory safety before attempting the detailed methods. No turn-key software or hardware distribution was verified.","updated_at":"2026-10-17T15:13:29Z","url":"https://research.wusong.net/cases/autonomous-scanning-droplet-cell-alloy-design","sources":[{"label":"Towards automated design of corrosion resistant alloy coatings with an autonomous scanning droplet cell","url":"https://arxiv.org/abs/2203.17049","kind":"arxiv","is_primary":1}]},{"slug":"autora","title":"AutoRA","summary":"AutoRA connects experimental design, synthetic or real data collection, and model fitting through shared state. The peer-reviewed 2024 software article describes version 4.2.0 and configurable components, rather than an inherently language-model-based agent or a new efficacy benchmark. Researchers must configure the study and assess models and results. This review did not reproduce a participant study or scientific finding.","resource_type":"framework","disciplines":["psychology"],"tasks":["experiment-design","workflow-orchestration"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"L2 applies only to the intended human-configured closed-loop empirical-research architecture. Users define the empirical problem and choose experimental-design, collection and theory-fitting components. A shared state connects selection, observations and fitting; the software article does not establish a particular participant study or universally enforced consent/approval controls. Termination and approvals depend on the user-defined orchestration/components; no universal agent-level or empirical efficacy claim.","source_checked_date":"2026-10-19","limitations":"Packages/components must be configured; hidden component state can defeat reproducibility. Real-world runners may involve recruitment/cloud storage. Ethics, participant consent, model validity and enforced human checkpoints are not guaranteed by this article. JOSS software review is not independent scientific replication.","adoption_requirements":"Untested suggestion: inspect the pinned sources and validate data, evaluator assumptions, implementation availability and researcher controls before planning a bounded trial. No setup or trial was executed.","updated_at":"2026-10-20T08:01:56Z","url":"https://research.wusong.net/cases/autora","sources":[{"label":"Primary manuscript — JOSS9(104),6839;published5December2024;paper describes AutoRA4.2.0","url":"https://joss.theoj.org/papers/10.21105/joss.06839.pdf","kind":"doi","is_primary":1},{"label":"Primary project repository or artifact (version limits apply)","url":"https://github.com/AutoResearch/autora","kind":"code-repository","is_primary":1}]},{"slug":"awesome-literature-review-skill","title":"awesome-literature-review: narrative review drafting skill","summary":"Claude Code skill documented for four-part narrative literature-review drafts, with topic and search-scope alignment before drafting, user checking of the literature list and evidence-level restrictions on claims. The documentation distinguishes narrative review writing from systematic-review screening and meta-analysis.","resource_type":"agent-skill","disciplines":["multidisciplinary"],"tasks":["literature-research","scientific-writing"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation unspecified. The researcher defines the task, controls access and checks sources and outputs.","source_checked_date":"2026-10-17","limitations":"README-only review; no skill execution, citation checking or document-format validation reproduced. Prompt instructions do not guarantee factual compliance. Explicitly not a substitute for systematic-review screening, risk-of-bias evaluation or meta-analysis. Sponsorship acknowledgements are not evidence of research quality.","adoption_requirements":"Compatible skill host, SKILL.md and references directory, accessible source records and researcher oversight are required. Confirm database access rights and document privacy; check cited content rather than trusting draft fluency.","updated_at":"2026-10-17T17:46:50Z","url":"https://research.wusong.net/cases/awesome-literature-review-skill","sources":[{"label":"yzbbds520/awesome-literature-review — primary README","url":"https://github.com/yzbbds520/awesome-literature-review","kind":"code-repository","is_primary":1}]},{"slug":"awesomecua-research-catalog","title":"AwesomeCUA: computer-use agent research catalog","summary":"AwesomeCUA documents a source-linked catalog of computer-use, GUI, browser and mobile agents, with research-paper discovery and JSONL/SQLite metadata downloads for people, RAG systems and agents. The repository is the static website and metadata snapshot, not the underlying models or datasets.","resource_type":"tool","disciplines":["computer-science"],"tasks":["data-management","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No automation level assigned. This catalog supports evidence discovery and agent ingestion; humans remain responsible for selecting sources and distinguishing bibliographic metadata from validated research.","source_checked_date":"2026-10-17","limitations":"README-only review: the website, download contents, individual papers, coverage claims and automation runs were not independently checked. No catalog completeness, annotation accuracy, retrieval quality or benchmark result is established. Upstream claims remain unverified by this review; metadata-only records must not be treated as reviewed evidence.","adoption_requirements":"README describes a static site requiring no account or API key for browsing. Agent/RAG reuse still requires source-rights review, filtering by review status and checking upstream licenses. Downloaded metadata and annotations are not upstream training data, weights or papers.","updated_at":"2026-10-17T21:43:09Z","url":"https://research.wusong.net/cases/awesomecua-research-catalog","sources":[{"label":"AwesomeCUA: computer-use agent research catalog — repository README","url":"https://github.com/AwesomeCUA/awesomecua.github.io","kind":"code-repository","is_primary":1}]},{"slug":"awesomelit-paper-literature-hypotheses","title":"AwesomeLit: user-steered literature exploration and hypothesis framing","summary":"AwesomeLit is described as a human-agent collaborative visual system for exploring unfamiliar literature and framing possible research directions.","resource_type":"use-case","disciplines":["multidisciplinary"],"tasks":["hypothesis-generation","literature-research"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":"L2","workflow_note":"Scoped L2 for the abstract-described, user-steerable literature process: researcher intentions guide query-tree exploration and paper-similarity inspection towards possible research directions. Detailed planning, tool use, human validation and bounded experiments needed for L3/L4 are not established. Qualitative study findings remain author-reported.","source_checked_date":"2026-10-19","limitations":"Official abstract only. No full paper, software, implementation or study instrument was independently checked. Transparency, provenance, novelty and effectiveness are author descriptions, not verified controls. Reported confidence or usefulness does not establish discovery, hypothesis feasibility or reliable gap detection. No paper-to-repository linkage or current deployment is inferred.","adoption_requirements":"Untested suggestion: inspect the original paper and study limits; independently check literature coverage, source provenance and proposed directions before reuse.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/awesomelit-paper-literature-hypotheses","sources":[{"label":"Primary preprint abstract","url":"https://arxiv.org/abs/2603.22648","kind":"arxiv","is_primary":1}]},{"slug":"benchmarking-self-driving-labs-af-ef","title":"Benchmarking self-driving labs with acceleration and enhancement factors","summary":"A benchmark-design reference for machine-learning-guided self-driving laboratories, comparing acceleration factor (AF) and enhancement factor (EF) relative to a reference strategy.","resource_type":"workflow-process","disciplines":["chemistry","materials-science"],"tasks":["agent-evaluation","data-analysis","optimization"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level workflow automation intentionally unspecified. A reusable component/process reference is not a measured autonomous workflow; researcher oversight remains necessary.","source_checked_date":"2026-10-17","limitations":"Primary abstract plus a literature-survey opening excerpt for the linked artifact identity; official RSC abstract/publication metadata and publisher-deposited Crossref record reviewed. No equations, underlying literature corpus, benchmark dataset or simulation code was audited or executed. Reported median AF6 and EF peaks are omitted rather than generalized. First online publication is2025-10-09, while the journal issue is2026; source year2025 retains first arXiv release.","adoption_requirements":"Editorial, untested: specify a goal, experimental budget and appropriate reference strategy, and obtain the full metric definitions before using the comparison. Inspect any linked code/data and permissions separately; their existence is not evidence of validated compatibility or outcomes.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/benchmarking-self-driving-labs-af-ef","sources":[{"label":"Benchmarking Self-Driving Labs","url":"https://arxiv.org/abs/2508.06642","kind":"arxiv","is_primary":1},{"label":"Primary publication/identity evidence for Q114 (114-publication-source.json)","url":"https://pubs.rsc.org/en/content/articlelanding/2026/dd/d5dd00337g","kind":"doi","is_primary":1}]},{"slug":"bertopic","title":"BERTopic","summary":"BERTopic uses transformer embeddings and class-based TF-IDF for topic modelling of document collections.","resource_type":"tool","disciplines":["business","humanities","social-science"],"tasks":["data-analysis"],"evidence_level":"project-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Topic clusters are analytical aids, not validated interpretive themes by themselves. Embedding choice, language coverage, outliers and preprocessing require sensitivity checks; no effectiveness claim is reproduced.","adoption_requirements":"Python environment, a lawful text corpus and expertise to interpret and validate topic groupings.","updated_at":"2026-10-16T14:58:30Z","url":"https://research.wusong.net/cases/bertopic","sources":[{"label":"BERTopic - Maarten Grootendorst","url":"https://maartengr.github.io/BERTopic/index.html","kind":"project-documentation","is_primary":1}]},{"slug":"bibcheck-de-otio","title":"bibcheck: bibliography checks and human claim-support triage","summary":"A documented bibliography checker verifies identifier/existence/metadata and citation-key linkage, with an explicit human worklist for claim support it does not check.","resource_type":"tool","disciplines":["humanities","multidisciplinary"],"tasks":["literature-research","scientific-writing"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Intended L2 verification component of AI-assisted scholarly writing: bibliography/citation inputs pass deterministic identifier/existence/metadata/linkage checks before human quotation/claim-support triage. The checker need not invoke AI. Existence and metadata agreement do not prove source support.","source_checked_date":"2026-10-19","limitations":"Full README only; no source code, database requests, bibliography upload, test execution or linked docs audit. Verified is limited to existence/metadata agreement; no quotation wording or claim-support verification. Maintainer calls absence a fabrication signal, but incompleteness/lookup failures can require human investigation. Network required, requests may expose citation metadata/URLs to configured services; no privacy guarantee. Version/format and stable-interface claims are documentation, not compatibility verification. AI editorial review only; no source code, models or installation executed, no production/status/publication changes.","adoption_requirements":"README specifies Node20+ and a live internet connection; inspect configuration, database coverage, input rights and request handling before a separate trial. No package installed or bibliography supplied.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/bibcheck-de-otio","sources":[{"label":"de-otio/bibcheck primary-readme","url":"https://raw.githubusercontent.com/de-otio/bibcheck/a8824a5e33b0683c5660408d5729aef3bf5f247f/README.md","kind":"code-repository","is_primary":1}]},{"slug":"biochatter","title":"BioChatter","summary":"Backend framework connecting biomedical applications and BioCypher knowledge graphs to conversational AI.[1]","resource_type":"framework","disciplines":["biomedicine"],"tasks":["data-analysis","knowledge-synthesis"],"evidence_level":"project-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Framework capability is not clinical validation; biomedical outputs require expert checking. Documentation review only - no software execution or independent performance validation.","adoption_requirements":"Python application; model/backend configuration; domain data or knowledge graph.","updated_at":"2026-10-17T06:33:49Z","url":"https://research.wusong.net/cases/biochatter","sources":[{"label":"BioChatter - official documentation","url":"https://biochatter.org/","kind":"project-documentation","is_primary":1}]},{"slug":"biodiscoveryagent","title":"BioDiscoveryAgent","summary":"BioDiscoveryAgent iteratively selects genetic perturbations using earlier screening outcomes and optional literature, database and critic tools. The authors evaluate it by replaying existing genetic screens, with dataset-specific batch sizes. This supports a documented experiment-selection workflow; it does not demonstrate an autonomous wet laboratory. Code is public, but this review did not run it.","resource_type":"framework","disciplines":["biology"],"tasks":["experiment-design","target-discovery"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":"L3","workflow_note":"L3 applies only to the documented tool-assisted retrospective experiment-selection variant. Researchers define the phenotype, historic screen and fixed round/batch budgets, and assess recovered hits against the measured screen. Optional literature/database/critic tools and previous measured responses inform later selections. The retrospective data lookup is not a newly executed wet-lab experiment. Evaluated replay uses five rounds and dataset-specific batch sizes. Scope does not imply all baselines/ablations used identical optional tools or claim prospective wet-lab autonomy.","source_checked_date":"2026-10-19","limitations":"Retrospective replay of existing screens, including one unpublished dataset; not an autonomous wet-lab demonstration or independent replication.","adoption_requirements":"Untested suggestion: inspect the pinned sources and validate data, evaluator assumptions, implementation availability and researcher controls before planning a bounded trial. No setup or trial was executed.","updated_at":"2026-10-20T08:01:56Z","url":"https://research.wusong.net/cases/biodiscoveryagent","sources":[{"label":"Primary manuscript — 2405.17631v3","url":"https://arxiv.org/html/2405.17631v3","kind":"arxiv","is_primary":1},{"label":"Primary project repository or artifact (version limits apply)","url":"https://github.com/snap-stanford/BioDiscoveryAgent/blob/19673c7371c542cfa155ffaa9047c53525f9fe2e/README.md","kind":"code-repository","is_primary":1}]},{"slug":"biology-literature-review-notebook-abdullahi","title":"Biology literature-review notebook: AI query, extraction and synthesis","summary":"A public biology mini-project documents question-to-PubMed retrieval, model-assisted evidence extraction and synthesis, plus exploratory analysis code.","resource_type":"tutorial-template","disciplines":["biomedicine"],"tasks":["data-analysis","data-extraction","knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Intended L2 for question/date-scoped query generation, PubMed retrieval, structured extraction, CSV/Markdown synthesis and separately scripted analysis. User-defined scope is documented; expert extraction/citation and statistical validation are not. No validated clinical meta-analysis or autonomous experiment loop is claimed.","source_checked_date":"2026-10-19","limitations":"Full README plus selected notebook markdown/method-identifying code-source lines read; not full implementation or notebook outputs. The raw notebook is preserved but not executed. Clinical values/correlations/PMID narratives were not checked and are omitted from recommendations; different narrative sections give inconsistent correlation/aggregate interpretations. Do not treat abstract extraction and heterogeneous cohort summaries as validated pooled meta-analysis, causal trend or clinical advice. No patient data, literature database query, model endpoint or medical inference run. Moving default-branch sources are timestamped, not commit-pinned or atomically versioned. AI editorial review only; no source code, models or installation executed, no production/status/publication changes.","adoption_requirements":"Read the public notebook/template as untested educational documentation; any later clinical research use needs independent study/source/metric/extraction and methods review. No dependencies, corpus, credentials or notebooks run.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/biology-literature-review-notebook-abdullahi","sources":[{"label":"Saadullahi/Building-AI-Agents-for-Biology primary-readme","url":"https://raw.githubusercontent.com/Saadullahi/Building-AI-Agents-for-Biology/HEAD/README.md","kind":"code-repository","is_primary":1},{"label":"Saadullahi/Building-AI-Agents-for-Biology notebook-method-excerpt","url":"https://raw.githubusercontent.com/Saadullahi/Building-AI-Agents-for-Biology/HEAD/cell_free_DNA_methylation_based_Multi_Cancer_Early_Detection_assays.ipynb","kind":"project-documentation","is_primary":1}]},{"slug":"biomni","title":"Biomni","summary":"Authors describe a general-purpose biomedical agent combining retrieval-augmented planning with code execution and biomedical tools; systematic benchmarking is author-reported.[2]","resource_type":"use-case","disciplines":["biomedicine"],"tasks":["data-analysis","workflow-orchestration"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":"L3","workflow_note":"L3 reflects the authors' documented multi-step tool-using design; humans supervise and validate results, and reported outcomes remain author-reported.","source_checked_date":"2026-10-15","limitations":"The retrieved article is explicitly a preprint version; reported benchmarks are author-reported and were not independently reproduced, and a later journal version was not used as validation. Authors warn the public tool is not for protected health information. No software execution for this review.","adoption_requirements":"Configured agent environment, model access and appropriate biomedical datasets; expert oversight.","updated_at":"2026-10-17T06:33:49Z","url":"https://research.wusong.net/cases/biomni","sources":[{"label":"Biomni: a general-purpose biomedical AI agent (preprint, PMC12157518)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC12157518/","kind":"doi","is_primary":1}]},{"slug":"biosnn-plug","title":"BioSNN-Plug","summary":"A repository-documented spiking-neural-network research prototype exploring local learning rules and modality plugins. Its README describes a biosnn-bus skeleton and separate research implementations, but explicitly says the cognitive core and LLM collaboration have not started.","resource_type":"framework","disciplines":["computer-science"],"tasks":["experiment-design","research-coding"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level is assigned: this is a component research prototype, not demonstrated end-to-end LLM orchestration. Human researchers set hypotheses, run controlled comparisons and audit conclusions.","source_checked_date":"2026-10-17","limitations":"README-only review; no code, data, package release or experimental record was executed or independently audited. LLM orchestration and the cognitive core are explicitly unstarted, so this is not an operational autonomous scientist. No performance figures are adopted: the README contains corrections, failed criteria and a reference to 2026-10-18, later than this UTC retrieval date, whose chronology was not resolved. Claims of local learning do not mean pretrained encoders were trained without gradients. No production readiness is established.","adoption_requirements":"For the documented bus quickstart the README specifies Python 3.10+ and NumPy with no GPU; research lines have separate dependencies and dataset terms. Review the documented SpikingJelly OIOSL dependency restrictions separately from repository licensing, and use isolated non-sensitive test data.","updated_at":"2026-10-17T21:43:09Z","url":"https://research.wusong.net/cases/biosnn-plug","sources":[{"label":"BioSNN-Plug — repository README","url":"https://github.com/lzmd-arch/BioSNN-Plug","kind":"code-repository","is_primary":1}]},{"slug":"birdnet","title":"BirdNET","summary":"Open-source deep-learning acoustic identification and batch-analysis toolkit for biodiversity monitoring and citizen science.[6]","resource_type":"tool","disciplines":["biology"],"tasks":["data-analysis"],"evidence_level":"project-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Detections depend on recordings, habitat and species coverage and are screening signals, not validated occurrences. Documentation review only - no software execution or independent performance validation.","adoption_requirements":"Audio recordings; location/date metadata and ecological validation; Analyzer GUI or CLI.","updated_at":"2026-10-17T06:33:49Z","url":"https://research.wusong.net/cases/birdnet","sources":[{"label":"BirdNET - official project site (Cornell University)","url":"https://birdnet.cornell.edu/","kind":"project-documentation","is_primary":1}]},{"slug":"catalysis-ai-ml-flow-research-guidance","title":"Catalysis AI/ML workflows: expert-supervised flow-research guidance","summary":"A review connects AI/ML-supported flow-chemistry experimentation with expert-led interpretation.","resource_type":"framework","disciplines":["chemistry","engineering"],"tasks":["experiment-design","workflow-orchestration"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Intended L2 reference guidance connecting ML-driven flow experiments and feedback with an expert scientist in the active loop. This is not a newly implemented laboratory, an audited human gate or a documented L4 system.","source_checked_date":"2026-10-19","limitations":"Selected primary PDF excerpts only, obtained through potentially cached text extraction; raw PDF/version parity was not established. The whole paper, figures, cited systems, hardware, code and outcomes were not audited. Broad advocacy does not establish catalysis benefits or laboratory performance, and cited implementations retain their own evidence limits.","adoption_requirements":"Untested suggestion: inspect the original method and its limits, then plan independent source/output checks before research reuse.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/catalysis-ai-ml-flow-research-guidance","sources":[{"label":"Selected primary paper passages","url":"https://arxiv.org/pdf/2603.05526","kind":"arxiv","is_primary":1}]},{"slug":"causalml","title":"CausalML","summary":"CausalML provides machine-learning methods for uplift modelling and conditional average treatment-effect estimation from experimental or observational data.","resource_type":"tool","disciplines":["business","economics","social-science"],"tasks":["data-analysis"],"evidence_level":"project-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Documentation does not establish causal identification in a new dataset. Researchers must justify design assumptions and check sensitivity, overlap and model fit; no empirical benefit was evaluated here.","adoption_requirements":"Python environment, intervention/outcome/covariate data and causal-inference expertise.","updated_at":"2026-10-16T14:58:30Z","url":"https://research.wusong.net/cases/causalml","sources":[{"label":"About CausalML","url":"https://causalml.readthedocs.io/en/latest/about.html","kind":"project-documentation","is_primary":1}]},{"slug":"cellpose","title":"Cellpose: deep-learning assistance for microscopy cell segmentation","summary":"The original Cellpose study applies a generalist deep-learning model to segment cells in varied microscopy images, with a 3D extension and software for manual labeling and result curation.","resource_type":"use-case","disciplines":["biology","biomedicine"],"tasks":["data-analysis","image-analysis"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":"L1","workflow_note":"A bounded image-analysis task: the model proposes cell masks. Researchers select images, inspect and correct segmentation, and validate downstream biological conclusions. This is non-agentic task assistance, not autonomous research.","source_checked_date":"2026-10-15","limitations":"Source review used the public publisher abstract, code/data-availability sections and current repository documentation, not the paywalled full article. No independent replication was performed. Domain shift, image quality and downstream measurement validity still require checking. Current Cellpose-SAM/DINO releases differ from the original study.","adoption_requirements":"Microscopy images, domain expertise for mask quality control, a suitable Python environment and sufficient memory/compute; review code, model and dataset licenses independently.","updated_at":"2026-10-15T17:25:09Z","url":"https://research.wusong.net/cases/cellpose","sources":[{"label":"Stringer et al., Cellpose: a generalist algorithm for cellular segmentation, Nature Methods (2021; online 2020)","url":"https://www.nature.com/articles/s41592-020-01018-x","kind":"doi","is_primary":1},{"label":"Cellpose primary repository and README (current version, not the original evaluation)","url":"https://github.com/MouseLand/cellpose","kind":"code-repository","is_primary":1},{"label":"Cellpose repository LICENSE","url":"https://github.com/MouseLand/cellpose/blob/main/LICENSE","kind":"project-documentation","is_primary":1}]},{"slug":"chatgpt-parent-interview-thematic-analysis","title":"Researcher-guided ChatGPT thematic analysis of parent interviews","summary":"An AI-assisted qualitative health-research workflow uses ChatGPT-4 for researcher-guided coding, candidate themes and reporting, with analysts reviewing outputs against parent interview transcripts.","resource_type":"workflow-process","disciplines":["medicine","social-science"],"tasks":["data-analysis"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Researcher-led workflow: people set research questions, approve data access and adjudicate outputs; no autonomous end-to-end performance claim. Workflow level intentionally unspecified.","source_checked_date":"2026-10-17","limitations":"Methods and abstract sections of publisher HTML reviewed (paper-excerpt); not a comprehensive full-paper audit or independent replication. Small interview study; parent perceptions do not establish treatment efficacy. Authors report three fabricated quotations, requiring transcript-level verification. Claims of efficiency and reduced bias are not established by a controlled comparison in the inspected methods. Recruitment wording refers to n=10 participants while results clarify 10 interviews with 12 parents; no sample-size efficacy inference is made.","adoption_requirements":"Ethics approval and consent appropriate to the new study, de-identified transcripts, institution-approved handling of sensitive health data, model access and qualitative-method expertise. Publication of a workflow does not authorize uploading confidential interviews.","updated_at":"2026-10-17T19:24:37Z","url":"https://research.wusong.net/cases/chatgpt-parent-interview-thematic-analysis","sources":[{"label":"Parent Perspectives on Dietary and Nutraceutical Therapies in Autism Spectrum Disorder: A Qualitative Thematic Analysis Using a Large Language Model","url":"https://doi.org/10.14200/jrm.2026.0001","kind":"doi","is_primary":1}]},{"slug":"chemcrow","title":"ChemCrow","summary":"LLM chemistry package integrating expert tools for reasoning-intensive chemical tasks.[3]","resource_type":"tool","disciplines":["chemistry"],"tasks":["experiment-design","workflow-orchestration"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L3","workflow_note":"L3 reflects the documented multi-step tool-using design; humans supervise, safety-check and validate results.","source_checked_date":"2026-10-15","limitations":"Repository documentation checked, not experimental runs; reaction suggestions need expert safety and feasibility assessment. No software execution or independent performance validation.","adoption_requirements":"Python and model API credentials; default reaction tools require RXN4Chem access.","updated_at":"2026-10-17T06:33:49Z","url":"https://research.wusong.net/cases/chemcrow","sources":[{"label":"ChemCrow - public repository (ur-whitelab/chemcrow-public)","url":"https://github.com/ur-whitelab/chemcrow-public","kind":"code-repository","is_primary":1}]},{"slug":"chemma-organic-synthesis","title":"Chemma: LLM-assisted organic synthesis exploration","summary":"Chemma is a fine-tuned language model described as an assistant for organic synthesis, with reaction-yield prediction and active-learning-guided experimental exploration.","resource_type":"tool","disciplines":["chemistry"],"tasks":["experiment-design","optimization"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Resource-level workflow classification unspecified; an instantiated workflow depends on the research task, implementation and oversight. Researchers retain responsibility for validation and safe use.","source_checked_date":"2026-10-17","limitations":"Abstract-only review of arXiv v1. Model weights, training-data provenance, contamination controls, detailed baselines and laboratory protocol were not inspected. Claims of state-of-the-art performance and human-like chemical understanding are not adopted. Reported reaction success does not establish broad yield reliability or safe autonomous synthesis.","adoption_requirements":"Untested editorial guidance: obtain and inspect the actual implementation and model/data terms separately, involve trained chemists, define reaction-space constraints, and validate predictions under an approved experimental protocol. No downloadable weights or turnkey service is confirmed.","updated_at":"2026-10-17T15:13:29Z","url":"https://research.wusong.net/cases/chemma-organic-synthesis","sources":[{"label":"Large Language Models to Accelerate Organic Chemistry Synthesis — primary arXiv record","url":"https://arxiv.org/abs/2504.18340","kind":"arxiv","is_primary":1}]},{"slug":"chemprop","title":"Chemprop","summary":"Message-passing neural networks for molecular property prediction.[4]","resource_type":"tool","disciplines":["chemistry"],"tasks":["data-analysis","research-coding"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Evaluation must account for domain shift and data leakage; training and prediction quality were not reproduced. No software execution or independent performance validation.","adoption_requirements":"Molecular structures with property labels, Python training environment and validation splits.","updated_at":"2026-10-17T06:33:49Z","url":"https://research.wusong.net/cases/chemprop","sources":[{"label":"Chemprop - repository README (chemprop/chemprop)","url":"https://github.com/chemprop/chemprop/blob/master/README.md","kind":"code-repository","is_primary":1}]},{"slug":"chemputation-validation-workflow","title":"Chemputation-based literature validation workflow","summary":"LLM-based workflow that validates synthetic literature procedures: extraction of procedures and analytical data, translation to universal XDL code, simulation on hardware-specific setups and execution on an XDL-controlled robotic system.[12]","resource_type":"workflow-process","disciplines":["chemistry","computer-science"],"tasks":["data-extraction","experiment-execution"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans set objectives and validate results.","source_checked_date":"2026-10-16","limitations":"Preprint; not independently reproduced. Requires the group's XDL/chemputation hardware and software stack. No code availability verified.","adoption_requirements":"XDL-controlled robotic chemistry platform (Chemputer-class hardware) plus the group's software stack; not a standalone package.","updated_at":"2026-10-17T08:00:32Z","url":"https://research.wusong.net/cases/chemputation-validation-workflow","sources":[{"label":"Validation of the Scientific Literature via Chemputation Augmented by Large Language Models - arXiv:2410.06384","url":"https://arxiv.org/abs/2410.06384","kind":"arxiv","is_primary":1}]},{"slug":"cladd","title":"CLADD","summary":"RAG-empowered agentic system for drug discovery: multiple LLM agents retrieve from biomedical knowledge bases, contextualize query molecules and integrate evidence without domain-specific fine-tuning; official code available.[16]","resource_type":"tool","disciplines":["biomedicine","computer-science"],"tasks":["drug-repurposing","knowledge-synthesis"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans set objectives and validate results.","source_checked_date":"2026-10-16","limitations":"Author-reported evaluations; not independently reproduced. Repository availability verified (GitHub API); no software executed.","adoption_requirements":"Python code from the official repository (github.com/Genentech/CLADD); LLM access and biomedical knowledge bases required; no independent setup was performed.","updated_at":"2026-10-17T08:00:32Z","url":"https://research.wusong.net/cases/cladd","sources":[{"label":"RAG-Enhanced Collaborative LLM Agents for Drug Discovery - arXiv:2502.17506","url":"https://arxiv.org/abs/2502.17506","kind":"arxiv","is_primary":1}]},{"slug":"claude-light-laboratory-prototyping","title":"Claude-Light: prototyping laboratory automation and ML workflows","summary":"Claude-Light is a lightweight, remotely accessible instrument for prototyping laboratory automation and machine-learning workflows, using a REST API, Raspberry Pi control, an RGB LED and a photometer measuring ten spectral outputs.","resource_type":"tool","disciplines":["chemistry","education"],"tasks":["experiment-design","optimization","research-coding"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Workflow level intentionally unspecified for a reusable component or conceptual reference; researcher oversight and task-specific validation remain necessary.","source_checked_date":"2026-10-17","limitations":"Abstract-only review. No hardware build, software configuration, device access terms or experiment was inspected or executed. The abstract acknowledges LLM reproducibility, security and reliability risks. Described demonstrations are author-reported and do not validate deployment in a chemical laboratory.","adoption_requirements":"The described instrument combines Raspberry Pi control, REST access, an LED and a photometer. Build instructions, access permissions and costs must be established separately; LLM use needs its own reliability and security assessment.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/claude-light-laboratory-prototyping","sources":[{"label":"The Evolving Role of Programming and LLMs in the Development of Self-Driving Laboratories","url":"https://arxiv.org/abs/2504.13870","kind":"arxiv","is_primary":1}]},{"slug":"claw-ai-lab-interactive-research-team","title":"Claw AI Lab: interactive multi-agent research platform","summary":"Claw AI Lab is an author-described interactive multi-agent research platform with configurable team roles, monitoring, artifact inspection and rollback/resume controls in a unified dashboard.","resource_type":"framework","disciplines":["computer-science"],"tasks":["experiment-execution","research-coding","scientific-writing","workflow-orchestration"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Resource-level automation unspecified; researchers select the task, assess data rights and inspect outputs. Source review is not an executed workflow.","source_checked_date":"2026-10-17","limitations":"Abstract-only arXiv v1 review; the author-linked repository was not inspected or installed. The internal evaluation uses five AI research case studies and AI expert judges against AutoResearchClaw. Judge configuration, case selection, code and raw outcomes were not inspected. Reported preferences do not establish independent novelty, research correctness or general reliability.","adoption_requirements":"Editorial, untested: inspect the official repository and complete methods for model, compute, sandbox and dataset requirements. Confirm rights to local datasets and protect credentials; supervise experiments and independently verify manuscript claims.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/claw-ai-lab-interactive-research-team","sources":[{"label":"Claw AI Lab: An Autonomous Multi-Agent Research Team","url":"https://arxiv.org/abs/2605.22662","kind":"arxiv","is_primary":1}]},{"slug":"clawxiv-signed-research-archive","title":"ClawXiv: signed human–AI research archives","summary":"Toolchain for turning human–AI co-research sessions into durable signed archival artifacts, pairing a readable publication with a content-addressed bundle.","resource_type":"workflow-process","disciplines":["multidisciplinary"],"tasks":["data-management","scientific-writing","workflow-orchestration"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Workflow level unspecified for this resource; researchers define objectives, control input data and validate outputs. Suggested adoption steps were not exercised.","source_checked_date":"2026-10-17","limitations":"ArXiv abstract-only plus repository documentation review. Source documentation only; no bundle, signing, publication or dependency setup tested. Cryptographic signing does not establish scientific correctness or independently authenticate all claimed AI authorship. The paper mentions a content-safety stub, not a validated safety filter. Extraction returned an arXiv v1 record; queued v2 version not independently inspected.","adoption_requirements":"Existing research materials and local dependency toolchain; appropriate signing-key custody and publication-service access. Inspect configure and Makefile documentation; deployment requirements were not independently tested.","updated_at":"2026-10-17T11:14:58Z","url":"https://research.wusong.net/cases/clawxiv-signed-research-archive","sources":[{"label":"[2604.16476] ClawXiv: a signed archival workflow and distributed publication architecture for human--AI collaborative research","url":"https://arxiv.org/abs/2604.16476","kind":"arxiv","is_primary":1},{"label":"GitHub - kornai/clawxiv: A public, distributed, author-signed preprint archive for human--AI collaborative research","url":"https://github.com/kornai/clawxiv","kind":"code-repository","is_primary":1}]},{"slug":"clinical-transcript-thematic-analysis-evaluation","title":"Evaluation framework for LLM thematic analysis of clinical transcripts","summary":"Position paper proposing validity, reliability and interpretability as evaluation dimensions for LLM-assisted thematic analysis of clinical transcripts.","resource_type":"framework","disciplines":["medicine","social-science"],"tasks":["data-analysis"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Method/framework-level autonomy is unspecified; researchers choose evaluation criteria, govern data and interpret outputs.","source_checked_date":"2026-10-17","limitations":"A position and evaluation framework, not a validated clinical analysis tool or demonstrated patient benefit. Workshop presentation is an arXiv comment, not independently verified peer-review status. Primary-source abstract and metadata review only; full-text methods, implementation and data were not inspected. No model or experiment was run and no outcome independently reproduced.","adoption_requirements":"Use as a methodological reference; applying it requires qualitative and clinical expertise, an evaluation protocol and appropriate governance for sensitive transcripts. No software deployment was inspected.","updated_at":"2026-10-17T11:14:58Z","url":"https://research.wusong.net/cases/clinical-transcript-thematic-analysis-evaluation","sources":[{"label":"Position: Thematic Analysis of Unstructured Clinical Transcripts with Large Language Models — arXiv","url":"https://arxiv.org/abs/2509.14597","kind":"arxiv","is_primary":1}]},{"slug":"co-scientist","title":"Co-Scientist: accelerating scientific discovery with a multi-agent AI system","summary":"A Gemini-based multi-agent system for structured scientific thinking and hypothesis generation that continuously generates, critiques and refines hypotheses, scaling test-time compute. Across three biomedical applications - drug repurposing, novel-target discovery, and mechanisms of antimicrobial resistance - it identified drug-repurposing candidates and synergistic combination therapies for acute myeloid leukaemia that were validated in vitro.","resource_type":"use-case","disciplines":["biomedicine","medicine"],"tasks":["drug-repurposing","hypothesis-generation","mechanism-elucidation","target-discovery"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":"L3","workflow_note":"Coordinated agents generate, critique and refine hypotheses using evidence within human-defined objectives. Humans perform experimental validation; the documented scope is not autonomous execution of an entire research cycle.","source_checked_date":"2026-10-15","limitations":"Validated examples are three focused biomedical applications; general scientific utility beyond these is not established. Validation is in vitro. Hospital-partner involvement means the work is not reproducible outside similar collaborations without the same data access. Not independently reproduced.","adoption_requirements":"Access to the underlying frontier model series (Gemini) and sufficient test-time compute; close collaboration with a wet-lab or clinical partner capable of validating generated hypotheses; domain experts to formulate the research objective precisely.","updated_at":"2026-10-15T17:25:09Z","url":"https://research.wusong.net/cases/co-scientist","sources":[{"label":"Accelerating scientific discovery with Co-Scientist, Nature (2026)","url":"https://doi.org/10.1038/s41586-026-10644-y","kind":"doi","is_primary":1}]},{"slug":"codebook-reduction-and-saturation","title":"Codebook Reduction and Saturation","summary":"Methodological paper on assessing saturation of LLM-generated initial qualitative codes, proposing a DSPy-based technique for measuring Inductive Thematic Saturation.","resource_type":"workflow-process","disciplines":["social-science"],"tasks":["data-analysis"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Method/framework-level autonomy is unspecified; researchers choose evaluation criteria, govern data and interpret outputs.","source_checked_date":"2026-10-17","limitations":"The abstract asserts precise saturation measurement but does not establish independent validity or equivalence to interpretive qualitative analysis. Primary-source abstract and metadata review only; full-text methods, implementation and data were not inspected. No model or experiment was run and no outcome independently reproduced.","adoption_requirements":"For adaptation, qualitative-method expertise, interview data with appropriate permissions, LLM access and DSPy familiarity are likely prerequisites; exact implementation requirements were not inspected.","updated_at":"2026-10-17T11:14:58Z","url":"https://research.wusong.net/cases/codebook-reduction-and-saturation","sources":[{"label":"Codebook Reduction and Saturation: Novel observations on Inductive Thematic Saturation for Large Language Models and initial coding in Thematic Analysis — arXiv","url":"https://arxiv.org/abs/2503.04859","kind":"arxiv","is_primary":1}]},{"slug":"codepulse-py-research-process","title":"CodePulse-Py: an early LLM-assisted programming-education research process","summary":"Early research repository for studying LLM-assisted Python misconception detection, documenting reconstruction of student code states from keystroke logs; its app services are explicitly stubs.","resource_type":"workflow-process","disciplines":["computer-science","education"],"tasks":["data-analysis","data-management","research-coding"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation left unspecified. Researchers/educators choose materials, govern model access and verify interpretations and outputs.","source_checked_date":"2026-10-17","limitations":"Primary README plus data/replay documentation only; no code audit or execution. The README explicitly says application services are health-check stubs: this listing is an early research process/prototype, not a functioning adaptive tutor or verified fine-tuned model. Replay checks cover syntax, not exact code recovery. Source datasets are external, have coverage gaps and may retain identifying keystrokes; use their actual terms and protect students. No misconception accuracy, intervention benefit or reported replay performance endorsed.","adoption_requirements":"Python 3.11, documented dependencies and separately obtained Edwards CS1 keystroke releases. Respect dataset access/usage terms and residual personal-data risks. Docker/Compose instructions launch documented stub services, not a finished product.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/codepulse-py-research-process","sources":[{"label":"Codepulse-Py/codepulse-py — primary README","url":"https://github.com/Codepulse-Py/codepulse-py","kind":"code-repository","is_primary":1},{"label":"CodePulse-Py — data card","url":"https://raw.githubusercontent.com/Codepulse-Py/codepulse-py/main/docs/data.md","kind":"project-documentation","is_primary":1},{"label":"CodePulse-Py — replay methodology","url":"https://raw.githubusercontent.com/Codepulse-Py/codepulse-py/main/docs/replay.md","kind":"project-documentation","is_primary":1}]},{"slug":"codex-prism","title":"Codex Prism","summary":"Local-first writing workspace that combines Codex assistance, LaTeX and PDF preview so scientific documents can be drafted, typeset and checked in a single environment.","resource_type":"tool","disciplines":["computer-science","multidisciplinary"],"tasks":["scientific-writing"],"evidence_level":"repository-documented","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Single-workspace loop for drafting: Codex-assisted text, LaTeX typesetting and PDF preview side by side.","source_checked_date":"2026-10-17","limitations":"Author-described tooling; assessment limited to repository documentation; not executed during this review.","adoption_requirements":"Installed Codex CLI and its access configuration; README advertises Ubuntu 26.04 LTS x86_64 packages. Consult setup documentation for LaTeX/toolchain requirements; MIT code does not establish zero provider costs.","updated_at":"2026-10-17T09:28:23Z","url":"https://research.wusong.net/cases/codex-prism","sources":[{"label":"codex-prism - GitHub repository (beniz)","url":"https://github.com/beniz/codex-prism","kind":"code-repository","is_primary":1}]},{"slug":"colinpetree-lit-review","title":"Lit Review: local scholarly search and AI abstract ranking","summary":"Locally run literature-search tool whose README describes searching OpenAlex, PubMed, Semantic Scholar and Scopus, then grading abstracts against a research question with a user-selected AI provider. ","resource_type":"tool","disciplines":["multidisciplinary"],"tasks":["literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"L2 applies to the documented connected process from a researcher-defined question through scholarly database search to chosen-model abstract grading and relevance ranking. The researcher chooses services/model and controls the run limit. The short README does not establish adaptive agent planning, a compulsory final-human-approval gate or a scientific propose/execute/evaluate loop. Checking ranking/eligibility before research use is an editorial recommendation, not a verified built-in control. Local storage does not mean local inference; no source software or outcomes were reproduced.","source_checked_date":"2026-10-17","limitations":"README-only review; releases, installation, encryption, cost controls and retrieval/ranking quality were not tested. The README describes mandatory version checks and external transmission of searches and abstracts; this is not a fully offline or no-egress tool. Researcher judgment is needed for relevance and coverage.","adoption_requirements":"A compatible documented Windows or Mac build, scholarly database access as required, and the user's own AI provider key and budget. Review provider terms and permission to transmit abstracts; local storage does not mean offline inference.","updated_at":"2026-10-17T21:22:00Z","url":"https://research.wusong.net/cases/colinpetree-lit-review","sources":[{"label":"colinpetree/lit-review primary documentation","url":"https://github.com/colinpetree/lit-review","kind":"code-repository","is_primary":1}]},{"slug":"collaborative-graph-learning-design-map","title":"Collaborative graph learning: a conceptual research design map","summary":"A conceptual framework surveying collaborative learning from Euclidean to graph-structured data, organized around learning effectiveness, efficiency and privacy, and distinguishing independent graph instances from partitions of a shared graph.","resource_type":"framework","disciplines":["computer-science"],"tasks":["data-analysis","knowledge-synthesis"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Workflow level intentionally unspecified for a reusable component or conceptual reference; researcher oversight and task-specific validation remain necessary.","source_checked_date":"2026-10-17","limitations":"Abstract plus HTML introduction and section map reviewed; later algorithms, equations and appendices were not substantively read. Official OpenReview publication metadata was additionally checked by an independent AI audit, but referee reports and the collaborative_gnns repository were not inspected. The introduction limits the survey to supervised learning. This is a conceptual design map, not a new implementation or privacy guarantee.","adoption_requirements":"Editorial, untested use requires identifying graph ownership/partition structure, collaboration constraints and privacy obligations. Select and validate an implementation separately; keeping data local is not by itself a privacy proof.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/collaborative-graph-learning-design-map","sources":[{"label":"From Euclidean to Graph-Structured Data: A Survey of Collaborative Learning","url":"https://arxiv.org/abs/2609.02984","kind":"arxiv","is_primary":1},{"label":"Primary HTML introduction","url":"https://arxiv.org/html/2609.02984","kind":"arxiv","is_primary":1},{"label":"Official TMLR OpenReview publication metadata","url":"https://openreview.net/forum?id=vj9l8AjLT6","kind":"project-documentation","is_primary":1}]},{"slug":"compressing-validation-bottleneck-proposed-sdl","title":"Compressing the Validation Bottleneck: proposed agentic SDL methods","summary":"An arXiv abstract proposes combining prior-aware experimental design and uncertainty-based measurement-cost selection to reduce self-driving-lab validation bottlenecks.","resource_type":"workflow-process","disciplines":["biology","materials-science"],"tasks":["experiment-design","workflow-orchestration"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level assigned. The abstract proposes prior-aware experiment selection and uncertainty-based choice between high- and low-cost measurements. Researcher approvals, execution boundaries and stopping conditions are not specified; an aim to accelerate research is not demonstrated L4 operation or a measured saving.","source_checked_date":"2026-10-19","limitations":"Official arXiv abstract only. The full paper, implementation, laboratory operation and measurement surrogate were not audited. The components are proposed; no completed system or empirical reduction is established. Workshop acceptance is author-supplied metadata, not independently verified peer review. No software, instrument or experiment was run.","adoption_requirements":"Untested suggestion: inspect the primary paper and establish implementation, data rights, measurement validity and laboratory-safety requirements before planning a trial.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/compressing-validation-bottleneck-proposed-sdl","sources":[{"label":"Primary preprint abstract","url":"https://arxiv.org/abs/2607.04508","kind":"arxiv","is_primary":1}]},{"slug":"computational-safety-hypothesis-testing","title":"Computational Safety for Generative AI: A Hypothesis Testing Perspective","summary":"A mathematical framework formulates generative-AI safety questions as hypothesis-testing problems, covering malicious-prompt detection and identification of AI-generated content.","resource_type":"framework","disciplines":["computer-science"],"tasks":["data-analysis"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Automation level unspecified: a resource-level description is not a measured autonomous workflow. Researchers select goals and data and validate outputs.","source_checked_date":"2026-10-17","limitations":"Primary arXiv abstract-only review; full text and software not reviewed or executed. No detection accuracy, operational reliability or deployment effectiveness is established by this abstract review. Equations, assumptions and experiments were not inspected; this is not a verified safety tool.","adoption_requirements":"Editorial, untested guidance: read the full framework before defining hypotheses and error criteria. Use appropriately permissioned data and researcher oversight; do not treat a detector result as proof of malicious intent or machine authorship.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/computational-safety-hypothesis-testing","sources":[{"label":"Computational Safety for Generative AI: A Hypothesis Testing Perspective — arXiv 2502.12445","url":"https://arxiv.org/abs/2502.12445","kind":"arxiv","is_primary":1}]},{"slug":"core-bench","title":"CORE-Bench","summary":"Benchmark evaluating whether agents can computationally reproduce published research artifacts.[14]","resource_type":"benchmark-dataset","disciplines":["multidisciplinary"],"tasks":["agent-evaluation"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"No code run; running unfamiliar publication code requires sandboxing. The original harness is explicitly no longer actively maintained. Documentation review only - no software execution or independent performance validation.","adoption_requirements":"Isolated execution environments, benchmark data and an agent/harness; potentially substantial compute.","updated_at":"2026-10-17T06:33:49Z","url":"https://research.wusong.net/cases/core-bench","sources":[{"label":"CORE-Bench - repository (siegelz/core-bench)","url":"https://github.com/siegelz/core-bench","kind":"code-repository","is_primary":1}]},{"slug":"coscientist","title":"Coscientist: autonomous chemical research with large language models","summary":"A GPT-4-driven system that plans and executes chemistry experiments across internet and documentation search, code execution, and laboratory automation. In the reported work it autonomously designed, planned and performed complex experiments across six diverse tasks, including reaction optimization of palladium-catalysed cross-couplings.","resource_type":"use-case","disciplines":["chemistry"],"tasks":["experiment-design","experiment-execution","optimization"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":"L4","workflow_note":"The documented chemistry workflow closes planning, tool use, execution and reaction optimization into an iterative loop. Human researchers define tasks, provision apparatus and must oversee chemical safety and validate results.","source_checked_date":"2026-10-15","limitations":"As reported, the system depends on a frontier proprietary model (GPT-4) and on specific automation hardware for the physical loop. Reported capability is demonstrated on a small set of tasks rather than a broad chemistry benchmark, and the authors' claims have not been independently reproduced.","adoption_requirements":"Frontier LLM API access; computational chemistry tooling; access to automated synthesis/analysis hardware for closed-loop experiments; a researcher able to validate that planned experiments are safe and chemically sound.","updated_at":"2026-10-15T17:25:09Z","url":"https://research.wusong.net/cases/coscientist","sources":[{"label":"Boiko et al., Autonomous chemical research with large language models, Nature (2023)","url":"https://doi.org/10.1038/s41586-023-06792-0","kind":"doi","is_primary":1},{"label":"Code repository (Coscientist)","url":"https://github.com/gomesgroup/coscientist","kind":"code-repository","is_primary":0}]},{"slug":"coti-qualitative-analysis","title":"CoTI: Collaborative Theme Identification Agent","summary":"Multi-agent framework supporting qualitative analysis of patient interviews through key-phrase extraction, theme identification and codebook generation.","resource_type":"framework","disciplines":["medicine","social-science"],"tasks":["data-analysis","data-extraction"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Workflow level unspecified for this resource; researchers define objectives, control input data and validate outputs. Suggested adoption steps were not exercised.","source_checked_date":"2026-10-17","limitations":"Abstract-level review of the retrieved v1 record only; queued metadata names a later title/version not returned by this extraction. Findings on 12 interviews and marginal collaboration gains are author-reported, not independently reproduced. Sensitive interview data require appropriate consent, privacy safeguards and expert interpretation.","adoption_requirements":"Review the method and establish model access, approved transcript handling and a domain-led coding protocol; distribution and full setup were not verified.","updated_at":"2026-10-17T11:14:58Z","url":"https://research.wusong.net/cases/coti-qualitative-analysis","sources":[{"label":"[2512.16063] A Multi-Agent Large Language Model Framework for Automated Qualitative Analysis","url":"https://arxiv.org/abs/2512.16063","kind":"arxiv","is_primary":1}]},{"slug":"critical-business-research-agent-skill","title":"Critical Research Agent: business-research guidance","summary":"A documented business-research skill prescribes contextual questions, five evidence pillars, source/confidence labels and approval of a research brief before creating a deck.","resource_type":"agent-skill","disciplines":["business","social-science"],"tasks":["data-analysis","knowledge-synthesis"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Intended L2 business-research guidance: context questions, five evidence pillars, source/confidence tagging and triangulation lead to a brief, followed by approval before a deck. These are instructions/checkpoints, not verified enforcement or anti-hallucination effectiveness.","source_checked_date":"2026-10-19","limitations":"Full short README read; no underlying skill/code/model execution, benchmark or actual client/academic study. Production-grade, forces, validates everything and anti-hallucination language are source claims/instructions, not verified controls/effectiveness. Agreement of multiple sources can retain shared bias. Bilingual capability is claimed, not tested. No financial, strategic or investment recommendation produced. AI editorial source review only; no installation, source code execution, provider/model calls or outcome reproduction.","adoption_requirements":"README names a Claude skill; precise installation, model/tool requirements and privacy/provider terms are not established by the checked README. Inspect original skill before any separately authorised trial.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/critical-business-research-agent-skill","sources":[{"label":"ap-artificial/critical-research-agent-skill primary-readme","url":"https://raw.githubusercontent.com/ap-artificial/critical-research-agent-skill/eb586f4f060632826f9e31532bf528cd4b9dfe83/README.md","kind":"code-repository","is_primary":1}]},{"slug":"crux-pi-approved-computational-research-loop","title":"Crux: PI-approved bounded computational research loops","summary":"A toolbox combining research bookkeeping and external-agent workflows. The assessed L4 scope is its documented PI-approved, budgeted single-objective autopilot, not the read-only cockpit or a universal autonomy claim about the whole framework.","resource_type":"framework","disciplines":["computer-science","multidisciplinary"],"tasks":["experiment-execution","hypothesis-generation","optimization","research-coding","workflow-orchestration"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L4","workflow_note":"Documented-loop L4 only for a PI-approved computational autopilot with a fixed objective, researcher scope/design controls and attempt/time/worker-call limits. PI approval delegates attempts; final synthesis, answers and merges remain human decisions. Not the deterministic checker or demonstrated autonomous science. Enforcement/outcomes were not executed or verified.","source_checked_date":"2026-10-18","limitations":"Documentation review, not implementation/security/enforcement audit. Ordinary notebook PI approval rules differ from the approved-autopilot exception; individual attempts need no new signature inside that remit. The general close path can warn about missing files, but later05.3 autopilot documentation writes and links a report for every attempt; the generic warning must not be treated as normal later-autopilot report omission. Earlier05.2 steward refusal does not generalise to later05.3 bounded steward/island delegation. Guidance is excluded from plan-hash coverage. Worker-invocation limits do not prove a financial/token cap, and fixed checks do not establish causal validity or prevention of selective analysis.","adoption_requirements":"A configured external coding/research agent, repository and suitable experiment/scorer, PI-chosen objective/null/thresholds, scoped writable/frozen paths and budget, plus genuine approval and validation processes. Documentation lists Python/git and optional host-specific tooling; dependencies/permissions and data rights need local assessment. No install or run was performed.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/crux-pi-approved-computational-research-loop","sources":[{"label":"Original announcement identifies Crux and explicitly links repository/website","url":"https://discuss.huggingface.co/t/crux-an-open-source-toolbox-for-agentic-computational-research/190714","kind":"community-discussion","is_primary":1},{"label":"README describes single-metric code/run/score attempts with PI direction/threshold/question decisions and prerequisites.","url":"https://github.com/mehdiforoozandeh/crux","kind":"code-repository","is_primary":1},{"label":"Autopilot preparation asks for PI design/harness/budget, creates a plan and stops before PI approval/run.","url":"https://raw.githubusercontent.com/mehdiforoozandeh/crux/9f7c945748deedcb582b95c390e3a80383fad0e7/skills/crux-autopilot/SKILL.md","kind":"project-documentation","is_primary":1},{"label":"Core instructions separate ordinary PI gates from delegated approved-autopilot attempts and reserve final synthesis/answers/merge","url":"https://raw.githubusercontent.com/mehdiforoozandeh/crux/9f7c945748deedcb582b95c390e3a80383fad0e7/skills/crux/SKILL.md","kind":"project-documentation","is_primary":1},{"label":"CLI documentation defines approval hash exceptions and explicit success/budget/abort/stall stopping conditions.","url":"https://raw.githubusercontent.com/mehdiforoozandeh/crux/9f7c945748deedcb582b95c390e3a80383fad0e7/skills/crux/scaffold/README.md","kind":"project-documentation","is_primary":1}]},{"slug":"cryo-et-object-identification-wayiok","title":"Cryo-ET Object Identification (wayiok)","summary":"Repository example organizing 2D object-detection and 3D segmentation approaches for cryo-electron tomography biological structures.","resource_type":"tutorial-template","disciplines":["biology"],"tasks":["image-analysis","research-coding"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Workflow level intentionally unspecified for this reusable resource. Researchers select inputs, validate outputs and retain responsibility for interpretation.","source_checked_date":"2026-10-17","limitations":"README and repository About review only; report, notebooks and datasets were not inspected or executed. README describes a YOLO11 2D approach but later calls its training code a 2D U-Net; this inconsistency is unresolved. No accuracy, comparative benefit, dataset completeness or reproducibility claim is validated.","adoption_requirements":"Requires ability to inspect and adapt machine-learning notebooks and appropriately licensed cryo-ET inputs. Exact dependencies, compute needs and data availability have not been verified.","updated_at":"2026-10-17T14:17:23Z","url":"https://research.wusong.net/cases/cryo-et-object-identification-wayiok","sources":[{"label":"wayiok/Cryo-electron-Tomography-Object-Identification primary repository documentation","url":"https://github.com/wayiok/Cryo-electron-Tomography-Object-Identification","kind":"code-repository","is_primary":1}]},{"slug":"currechat-helsinki","title":"CurreChat: university-controlled LLM chat and course-material RAG","summary":"University of Helsinki chat tool documented for education and research, with teacher-managed prompts and retrieval over supplied course materials.","resource_type":"tool","disciplines":["education","multidisciplinary"],"tasks":["knowledge-synthesis"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation left unspecified. Researchers/educators choose materials, govern model access and verify interpretations and outputs.","source_checked_date":"2026-10-17","limitations":"README-only review, not implementation or security audit. Institutional role-based access and a development setup are documented; availability outside Helsinki is not established. Uploaded files use S3 and OpenAI models use Azure, so confidentiality and retention need institutional review. No educational/research benefit, answer accuracy or privacy compliance evaluated.","adoption_requirements":"Documented development path requires npm, Docker/Compose and environment configuration, with PostgreSQL, Redis, Ollama, Azure and S3/MinIO services. Institutional access/data governance must be assessed.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/currechat-helsinki","sources":[{"label":"UniversityOfHelsinkiCS/gptwrapper — primary README","url":"https://github.com/UniversityOfHelsinkiCS/gptwrapper","kind":"code-repository","is_primary":1}]},{"slug":"ddi-ml-validation-review-framework","title":"Drug-interaction ML: representation and validation review framework","summary":"An Indonesian-language review abstract provides reference dimensions for assessing machine-learning drug-interaction prediction studies and their validation gaps.","resource_type":"framework","disciplines":["chemistry","computer-science","medicine"],"tasks":["experiment-design","literature-research"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level assigned. The review supplies nonnumeric reference dimensions for algorithm families, features, datasets, validation and metrics. AI in the prediction studies does not establish AI performing the review or one supervised AI research workflow. An archived PDF excerpt’s incompatible study-quality counts remain unresolved; comparative results are not accepted here.","source_checked_date":"2026-10-19","limitations":"Primary Indonesian abstract, selected journal policy and retained October 17 PDF excerpts only; not a fresh complete-PDF review, clinical data/model audit or reproduction of included studies. The abstract’s all-60-high-quality wording conflicts with a retained results figure of 63.3%; no pooled performance or quality-count conclusion is adopted. Inconsistent scope/licence boilerplate, drafting artefacts, unrelated footer links and differing policy reviewer counts limit provenance confidence; they do not prove compromise or misconduct. The journal-source label follows article identity/stated policy, not verified review conduct. No clinical impact, prescribing inference or drug-interaction advice is supplied.","adoption_requirements":"Untested suggestion: use only as a bounded checklist and verify full methods, study independence, leakage, external/cold-start validation and provenance before research reuse.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/ddi-ml-validation-review-framework","sources":[{"label":"Primary Indonesian article abstract","url":"https://doi.org/10.64036/pharmactive.v5i1.104","kind":"doi","is_primary":1},{"label":"Archived primary PDF: selected passages","url":"https://ejournal.bintangpersada.ac.id/index.php/jpharmactive/article/download/104/102","kind":"doi","is_primary":1}]},{"slug":"deep-learning-drug-repurposing-methods-databases","title":"Deep learning for drug repurposing: methods, databases, and applications","summary":"A methodological review introduces deep-learning drug-repurposing guidance, covering bioinformatics and pharmacogenomics databases, sequence- and graph-based representations, and COVID-19 applications.","resource_type":"tutorial-template","disciplines":["biomedicine"],"tasks":["drug-repurposing","knowledge-synthesis"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Automation level unspecified: a resource-level description is not a measured autonomous workflow. Researchers select goals and data and validate outputs.","source_checked_date":"2026-10-17","limitations":"Primary arXiv abstract-only review; full text and software not reviewed or executed. The abstract does not expose individual database recommendations, implementation steps or comparative outcomes. The record reports journal acceptance, but publisher full text was not reviewed. Examples do not establish clinical efficacy.","adoption_requirements":"Editorial, untested guidance: read the full review, then check current database access terms, data suitability and model validation requirements with biomedical specialists. Do not infer a ready-to-run pipeline or clinical recommendation.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/deep-learning-drug-repurposing-methods-databases","sources":[{"label":"Deep learning for drug repurposing: methods, databases, and applications — arXiv 2202.05145","url":"https://arxiv.org/abs/2202.05145","kind":"arxiv","is_primary":1}]},{"slug":"deepforest","title":"DeepForest","summary":"Open-source Python package training and predicting individual tree crowns from airborne RGB imagery, with a prebuilt NEON-trained model.[7]","resource_type":"tool","disciplines":["biology","earth-environmental"],"tasks":["image-analysis"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Detection accuracy varies with imagery and forest structure; the prebuilt model requires local annotation and retraining. Documentation review only - no software execution or independent performance validation.","adoption_requirements":"Georeferenced aerial imagery; Python; annotations for local adaptation.","updated_at":"2026-10-17T06:33:49Z","url":"https://research.wusong.net/cases/deepforest","sources":[{"label":"DeepForest - repository (weecology/DeepForest)","url":"https://github.com/weecology/DeepForest","kind":"code-repository","is_primary":1}]},{"slug":"deeplabcut","title":"DeepLabCut","summary":"Open-source deep-learning toolbox for markerless animal pose estimation supporting behavioural experiments.[12]","resource_type":"tool","disciplines":["biology","biomedicine"],"tasks":["image-analysis"],"evidence_level":"project-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Tracking quality and behavioural interpretation need domain-specific validation; no claim of universal animal accuracy. Documentation review only - no software execution or independent performance validation.","adoption_requirements":"Video data, visible landmarks and annotation/training or suitable pretrained model.","updated_at":"2026-10-17T06:33:49Z","url":"https://research.wusong.net/cases/deeplabcut","sources":[{"label":"DeepLabCut - official documentation (DeepLabCut project)","url":"https://deeplabcut.github.io/DeepLabCut/","kind":"project-documentation","is_primary":1}]},{"slug":"deepxde","title":"DeepXDE","summary":"Library for scientific machine learning and physics-informed learning, covering differential-equation and operator-learning problems.[10]","resource_type":"tool","disciplines":["engineering","physics"],"tasks":["data-analysis"],"evidence_level":"project-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Solver accuracy, physical consistency and convergence must be validated against domain baselines. Documentation review only - no software execution or independent performance validation.","adoption_requirements":"Specified equations, boundary/initial conditions or data; compatible ML backend.","updated_at":"2026-10-17T06:33:49Z","url":"https://research.wusong.net/cases/deepxde","sources":[{"label":"DeepXDE - official documentation (lululxvi/deepxde)","url":"https://github.com/lululxvi/deepxde","kind":"code-repository","is_primary":1}]},{"slug":"deer-flow","title":"DeerFlow 2.0: agent harness for research and coding","summary":"Open-source agent harness that composes sandboxes, memory, tools, subagents and a message gateway to run multi-step research, coding and creation tasks .","resource_type":"framework","disciplines":["computer-science"],"tasks":["knowledge-synthesis","literature-research","research-coding","workflow-orchestration"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"General orchestration framework; workflow level is unspecified until a concrete research workflow and oversight boundaries are defined.","source_checked_date":"2026-10-17","limitations":"Selected repository-documentation review only; no code audit, installation or outcome reproduction. Version 2.0 differs from the original deep-research framework. Code execution and file access require restricted deployment; provider costs/privacy were not assessed.","adoption_requirements":"Configure a supported LLM provider and execution/safety settings using the documented setup wizard. Editorial safeguards: assess provider privacy, restrict code/file access and review costs before supplying research data.","updated_at":"2026-10-17T14:17:23Z","url":"https://research.wusong.net/cases/deer-flow","sources":[{"label":"https://github.com/bytedance/deer-flow","url":"https://github.com/bytedance/deer-flow","kind":"code-repository","is_primary":1}]},{"slug":"dental-writing-ai-assisted-stylometric-analysis","title":"Dental writing: AI-assisted coding for stylometric research","summary":"A dental-writing study discloses Claude-assisted Python data processing within an author-directed retrospective linguistic analysis.","resource_type":"use-case","disciplines":["computer-science","linguistics","medicine"],"tasks":["data-analysis","research-coding"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":"L2","workflow_note":"Scoped L2 for the disclosed research-coding role: authors define the abstract corpus, features and comparisons; Claude assists Python parsing, feature computation and data organisation; authors retain scientific interpretation, conclusions and writing decisions. This role and the results are author-reported, not independently reproduced.","source_checked_date":"2026-10-19","limitations":"Selected primary HTML abstract, methods, AI-use statement, composite-index discussion and limitations only. The whole paper, figures, raw corpus, code, references and underlying model use were not audited. Six journals, two periods, abstract-only sampling, changing abstract length, editorial norms, language editing, demographics and possible human style adoption limit causal interpretation. The composite index lacks ground-truth validation. No individual AI-authorship detector, AI-adoption prevalence estimate, misconduct finding or clinical conclusion is established. Numerical drift does not prove authorship. Disclosed scripting assistance is not proof of code correctness or general model capability.","adoption_requirements":"Untested suggestion: prespecify corpus, features and baselines; verify parsing against source abstracts, retain researcher interpretation and require appropriate ground truth before individual classification.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/dental-writing-ai-assisted-stylometric-analysis","sources":[{"label":"Primary paper: selected methods, AI-use statement and limitations","url":"https://doi.org/10.1186/s41073-026-00226-x","kind":"doi","is_primary":1}]},{"slug":"derridai","title":"DerridAI","summary":"Local-first scholarly corpus environment documenting source ingestion, researcher review, provenance-aware metadata and evidence-grounded retrieval-augmented generation.","resource_type":"tool","disciplines":["humanities"],"tasks":["archival-research","data-management","knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Corpus setup, build, review and publication plus research querying form a defined researcher-supervised workflow; provenance architecture is documented, not independently tested here.","source_checked_date":"2026-10-17","limitations":"No installation, source-code security audit, retrieval/citation accuracy evaluation or scholarly outcomes reproduced. Local-first is not a blanket offline guarantee: imports and remote providers may use networks. Generated claims and automatically enriched metadata still require source checks. The retrieved README was head/tail truncated; reviewed introductory, architecture, setup and license sections support this bounded listing, not an exhaustive feature audit.","adoption_requirements":"Git, Docker with Compose, an LLM endpoint and configured generation/embedding models; optional local Ollama or other supported endpoint.","updated_at":"2026-10-17T11:14:58Z","url":"https://research.wusong.net/cases/derridai","sources":[{"label":"DerridAI — primary repository","url":"https://github.com/ajschlosser/DerridAI","kind":"code-repository","is_primary":1}]},{"slug":"dh-ai-african-studies-2026-archive-guidance","title":"DH/AI African Studies workshop archive: reference guidance for supervised archival processing","summary":"Workshop archive and collective-paper REFERENCE-GUIDANCE L2 for human-directed archival discovery/cataloguing: related OCR/HTR, summary and metadata pathways, expert verification and collection-specific purpose/rights/governance—not one prescribed serial architecture. It is not an archive-executed or empirically evaluated workflow, and cited project outcomes are not inherited.","resource_type":"workflow-process","disciplines":["humanities","social-science"],"tasks":["archival-research","data-extraction","data-management","knowledge-synthesis"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Reference-guidance L2 only for human-directed archival discovery/cataloguing through related, collection-specific OCR/HTR, summary/triage and metadata pathways. These are not one prescribed serial architecture or archive-executed workflow. Expert verification and purpose/rights/governance are normative; no universal ontology model, implemented safeguards or inherited project outcomes are claimed.","source_checked_date":"2026-10-19","limitations":"Paper-excerpt/section review of saved text, not a complete critical manuscript/PDF or publisher/DOI audit; repository metadata says corrected PDF unavailable. The guidance spans distinct projects and related pathways, not one prescribed sequential pipeline. The paper records NO CONSENSUS on balancing human-designed ontologies and AI-generated metadata; collaborative ontology design is not established as universal or achieved, and a participatory-design example reports skills/funding/power asymmetries. No archive-owned model/API sequence, execution trace, runtime approval/stop/budget enforcement, accuracy or actual outcome evaluation is established. Collection-specific rights/verification recommendations are not measured safeguards. No cited project benefits/results are inherited; private source notes were not accessed, and corrected public graph provenance is not a fresh private-note extraction.","adoption_requirements":"No installation is required to use the saved material as a reference. Any later application requires collection-specific choices about aims, tools, expertise and rights/governance, not proof of one universal chain or established collaborative categories. The paper recommends situated judgement, notes unresolved ontology balance, and supplies counterevidence to assuming equitable participation is achieved. Distinct documented code/site/paper/photograph terms remain separate. Editorial application advice is not a verified implementation feature.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/dh-ai-african-studies-2026-archive-guidance","sources":[{"label":"Original repository identifies the workshop archive and its collective scope.","url":"https://raw.githubusercontent.com/fmadore/dh-ai-african-studies-2026/76a086e91f114694b05eb478cb74cd2615b28db9/README.md","kind":"code-repository","is_primary":1},{"label":"Introduction describes human workshop discussion/collective drafting; this is not an archive runtime trace.","url":"https://raw.githubusercontent.com/fmadore/dh-ai-african-studies-2026/76a086e91f114694b05eb478cb74cd2615b28db9/src/lib/content/position-paper.md","kind":"project-documentation","is_primary":1},{"label":"Graph-script documentation describes deterministic curated-note-link extraction; private notes were not accessed.","url":"https://raw.githubusercontent.com/fmadore/dh-ai-african-studies-2026/76a086e91f114694b05eb478cb74cd2615b28db9/scripts/extract_concept_graph.py","kind":"project-documentation","is_primary":1},{"label":"Repository metadata supplies the paper publication/licence/PDF-availability declaration, not independent publisher verification.","url":"https://raw.githubusercontent.com/fmadore/dh-ai-african-studies-2026/76a086e91f114694b05eb478cb74cd2615b28db9/src/lib/data/position-paper-meta.ts","kind":"project-documentation","is_primary":1}]},{"slug":"discoverybench","title":"DiscoveryBench","summary":"Dataset and evaluation resources for data-driven discovery requiring statistical analysis and semantic reasoning.[15]","resource_type":"benchmark-dataset","disciplines":["multidisciplinary","social-science"],"tasks":["agent-evaluation","hypothesis-generation"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Benchmark results are not real-world scientific validity; real and synthetic datasets have distinct limitations. Documentation review only - no software execution or independent performance validation.","adoption_requirements":"Task datasets, evaluation harness and model/agent access; protected evaluation splits.","updated_at":"2026-10-17T06:33:49Z","url":"https://research.wusong.net/cases/discoverybench","sources":[{"label":"DiscoveryBench - repository (allenai/discoverybench)","url":"https://github.com/allenai/discoverybench","kind":"code-repository","is_primary":1}]},{"slug":"dml-llm-thematic-analysis","title":"LLM-Assisted Thematic Analysis Pipeline (Different Minds Lab)","summary":"Researcher-led Python pipeline for inductive thematic analysis of interview transcripts, with code extraction, code checking, theme generation and theme checking.","resource_type":"workflow-process","disciplines":["psychology","social-science"],"tasks":["data-analysis"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Human oversight at every stage is the documented intended workflow, not a claim that this editorial review or every pipeline run received human review.","source_checked_date":"2026-10-17","limitations":"Raw interviews are not released; derived de-identified outputs are documented. No runtime or time-saving claims reproduced, and conference presentation and lab affiliation are repository-reported. Model verification does not replace researcher judgment; sending participant text to an external API requires ethics/privacy review.","adoption_requirements":"Python virtual environment, requirements.txt, an OpenAI API key in the documented API_KEY setting, and appropriately consented transcripts.","updated_at":"2026-10-17T11:14:58Z","url":"https://research.wusong.net/cases/dml-llm-thematic-analysis","sources":[{"label":"LLM-Assisted Thematic Analysis Pipeline (Different Minds Lab) — primary repository","url":"https://github.com/jonathank256/LLMs-for-Thematic-Analysis","kind":"code-repository","is_primary":1}]},{"slug":"draft-detective","title":"Draft Detective","summary":"Open-source assistant that analyses a manuscript's citations — flagging unsupported or missing references and suggesting relevant literature — to make academic drafts more reviewable.","resource_type":"tool","disciplines":["computer-science","multidisciplinary"],"tasks":["automated-review","scientific-writing"],"evidence_level":"repository-documented","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Slots into pre-submission review: validate references, surface missing citations, strengthen the literature framing.","source_checked_date":"2026-10-17","limitations":"Author-described capabilities; accuracy and coverage were not independently assessed; no installation or evaluation executed.","adoption_requirements":"Supports Claude Code/Codex skills; saved README also shows a Python uv backend and pnpm frontend. Consult the selected interface setup for credentials and costs.","updated_at":"2026-10-17T09:28:23Z","url":"https://research.wusong.net/cases/draft-detective","sources":[{"label":"draft-detective - GitHub repository (agencyenterprise)","url":"https://github.com/agencyenterprise/draft-detective","kind":"code-repository","is_primary":1}]},{"slug":"drs-ml-validation-planning-checklist","title":"ML-assisted diffuse reflectance spectroscopy: validation-planning checklist","summary":"A review-derived conceptual checklist for planning ML-assisted diffuse-reflectance spectroscopy research: rigorous sample stratification, in-vivo validation and explainable algorithm development.","resource_type":"tutorial-template","disciplines":["biomedicine","computer-science"],"tasks":["data-analysis","experiment-design"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Workflow level intentionally unspecified for a conceptual reference. Researchers choose study designs, obtain appropriate approvals and validate conclusions; no autonomous clinical decision-making is proposed.","source_checked_date":"2026-10-17","limitations":"Abstract-only review of a 2025 systematic-review preprint. The reported 77-study corpus was not audited. No diagnostic model, detailed stratification protocol, in-vivo performance, clinical efficacy or executable pipeline was inspected or reproduced. The proposed checklist is an editorial, untested use of the authors' stated research priorities.","adoption_requirements":"Editorial, untested use: spectroscopy and ML expertise, lawful data access, an appropriate sample-stratification and validation plan, and ethics/safety review for any human-subject or in-vivo study. Detailed protocols, code and clinical suitability must be assessed separately.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/drs-ml-validation-planning-checklist","sources":[{"label":"Machine Learning Applications to Diffuse Reflectance Spectroscopy in Optical Diagnosis; A Systematic Review","url":"https://arxiv.org/abs/2503.02905","kind":"arxiv","is_primary":1}]},{"slug":"drugclip-contrastive-drug-disease-interaction","title":"DrugCLIP: Contrastive Drug-Disease Interaction For Drug Repurposing","summary":"DrugCLIP is a contrastive-learning method for drug-disease interaction modelling without negative labels, proposed for drug repurposing.","resource_type":"framework","disciplines":["biomedicine"],"tasks":["drug-repurposing"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Automation level unspecified: a resource-level description is not a measured autonomous workflow. Researchers select goals and data and validate outputs.","source_checked_date":"2026-10-17","limitations":"Primary arXiv abstract-only review; full text and software not reviewed or executed. Abstract-only evidence does not establish clinical efficacy, dataset release, reproducibility, negative sampling details or evaluation quality. The empirical evaluation is author-reported, with no numerical results verified.","adoption_requirements":"Editorial, untested guidance: inspect the methods, data rights and evaluation splits before implementation. Require biomedical expertise and independent validation; do not use model predictions as clinical advice or assume clinical-trial data may be redistributed.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/drugclip-contrastive-drug-disease-interaction","sources":[{"label":"DrugCLIP: Contrastive Drug-Disease Interaction For Drug Repurposing — arXiv 2407.02265","url":"https://arxiv.org/abs/2407.02265","kind":"arxiv","is_primary":1}]},{"slug":"e2er-end-to-end-research","title":"e2er (End-to-End Research)","summary":"Configurable AI-enabled research infrastructure organizing templates, specialist agents, reusable skills and researcher review pauses, with dossiers for checking study files and replication outputs.","resource_type":"framework","disciplines":["economics","multidisciplinary"],"tasks":["data-analysis","research-coding","workflow-orchestration"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Workflow level left unspecified for this resource-level listing. Researchers choose scope, inspect inputs and methods, and verify generated outputs before reuse.","source_checked_date":"2026-10-17","limitations":"Partial README review covering introduction, verification, replication, setup and study/review commands; linked dossiers and remaining documentation were not audited. File integrity is not scientific validity. Author replication examples and backend tests were not independently reproduced. The documented replication path excludes Stata and MATLAB; Windows and the Gemini backend are described as untested.","adoption_requirements":"Python 3.11–3.14 and supported AI access through a documented CLI subscription or API provider; supplied literature and data. The replication template also requires Docker. Review provider costs, permissions and data-disclosure risks; backend containment differs.","updated_at":"2026-10-17T14:30:44Z","url":"https://research.wusong.net/cases/e2er-end-to-end-research","sources":[{"label":"e2er (End-to-End Research) — primary repository README","url":"https://github.com/bhanneke/E2ER-project","kind":"code-repository","is_primary":1}]},{"slug":"econcslib-nikhil-garg","title":"EconCSLib (Nikhil Garg): AI-assisted paper formalization","summary":"Lean 4 library and AI-assisted workflow for formalizing economics and computation research papers, with agent skills and human-facing validation reports.","resource_type":"framework","disciplines":["computer-science","economics","mathematics"],"tasks":["proof-search","research-coding"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Workflow level unspecified for this resource; researchers define objectives, control input data and validate outputs. Suggested adoption steps were not exercised.","source_checked_date":"2026-10-17","limitations":"Paper extract and repository documentation only; no Lean build or proof audit executed. Lean checks formal statements under assumptions, not whether translation captures the original paper. Reported library size/completion is not independently checked; model access and costs vary.","adoption_requirements":"Lean/mathlib/CSLib toolchain, an LLM coding agent and source papers; repository specifies v4.30.0-rc2 at retrieval. Domain expertise remains necessary to inspect assumptions and translation fidelity.","updated_at":"2026-10-17T11:14:58Z","url":"https://research.wusong.net/cases/econcslib-nikhil-garg","sources":[{"label":"EconCSLib: AI-Assisted Lean Formalization for Economics & Computation Research","url":"https://arxiv.org/abs/2606.13306","kind":"arxiv","is_primary":1},{"label":"GitHub - nikhgarg/EconCSLib: AI-assisted Lean formalization for Economics and Computation research","url":"https://github.com/nikhgarg/EconCSLib","kind":"code-repository","is_primary":1}]},{"slug":"econml","title":"EconML: causal machine-learning methods for research","summary":"A Python framework that applies machine learning to individualized causal responses using observational or experimental data.","resource_type":"framework","disciplines":["business","economics","social-science"],"tasks":["data-analysis","research-coding"],"evidence_level":"project-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level entry: no universal autonomy level is assigned. The researcher defines the task, controls access and validates outputs; classify an actual instantiated workflow separately.","source_checked_date":"2026-10-15","limitations":"Flexible models do not eliminate confounding or identification assumptions. Documentation was reviewed, not a causal estimate, a code run or the claim that analysis becomes quicker.","adoption_requirements":"Python, treatment/outcome/covariate data and defensible causal identification assumptions. Access is open-source; the license was not reviewed.","updated_at":"2026-10-15T17:25:09Z","url":"https://research.wusong.net/cases/econml","sources":[{"label":"Reviewed primary source: EconML: causal machine-learning methods for research","url":"https://www.pywhy.org/EconML/spec/overview.html","kind":"project-documentation","is_primary":1},{"label":"Implementation linked by reviewed project documentation (not executed)","url":"https://github.com/py-why/econml","kind":"code-repository","is_primary":0}]},{"slug":"edubehaviors-assertion-annotation","title":"EduBehaviors: researcher-revised assertion schemas for educational dialogue","summary":"EduBehaviors decomposes educational-dialogue constructs into interpretable assertions and an explicit label-mapping rule, with researcher review and revision.","resource_type":"use-case","disciplines":["computer-science","education"],"tasks":["data-analysis","data-extraction"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":"L2","workflow_note":"Scoped L2 for researcher-defined constructs and approval/revision of generated assertions, LLM/encoder annotation, explicit mapping/classifier decisions and review-based iteration. Researchers judge assertion faithfulness and construct fit. Cross-model agreement does not establish construct validity or L3/L4 autonomy; TalkMoves results remain author-reported.","source_checked_date":"2026-10-19","limitations":"Selected original-v2 HTML abstract, framework/toolkit, performance, discussion and limitation sections only; not all appendices, references, code, data, dialogue, weights, toolkit/UI or model endpoints. The reported macro-F1 .764/kappa .756 is task-specific, using session-grouped folds over 50 TalkMoves sessions. Encoder results on 495 other sessions use a different setup and are not a common benchmark. Performance, cost and provider details were not independently verified. Shared LLM errors can survive agreement; assertion gold validation, out-of-context encoder limits and user discretion remain concerns. No general superiority, educational benefit, privacy, consent, de-identification or high-stakes grading guarantee is established. No student-level data was accessed.","adoption_requirements":"Untested suggestion: inspect the original v2 method, establish dialogue-data rights/privacy and construct-specific human gold checks, and assess transfer before a trial.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/edubehaviors-assertion-annotation","sources":[{"label":"Original v2 paper: selected framework and evaluation sections","url":"https://arxiv.org/html/2609.27043v2","kind":"arxiv","is_primary":1}]},{"slug":"effective-self-driving-laboratory-design-criteria","title":"Effective self-driving laboratories: speed and adaptability design criteria","summary":"A review-derived design reference proposing two criteria for an effective self-driving laboratory: outperform automation alone and adapt readily to new studies.","resource_type":"tutorial-template","disciplines":["materials-science"],"tasks":["experiment-design","workflow-orchestration"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Workflow level intentionally unspecified for a reusable component or conceptual reference; researcher oversight and task-specific validation remain necessary.","source_checked_date":"2026-10-17","limitations":"Publisher abstract and data-availability statement only; the substantive article body and detailed construction strategies were not available in the extraction. The abstract's acceleration framing is not a new experiment or a verified performance estimate. This is a conceptual design reference, not an executable laboratory recipe.","adoption_requirements":"Editorial, untested use: access the complete article and translate its speed and adaptability criteria into project-specific requirements with laboratory specialists. Hardware, cost and safety requirements were not read.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/effective-self-driving-laboratory-design-criteria","sources":[{"label":"How to build an effective self-driving laboratory","url":"https://link.springer.com/article/10.1557/s43577-023-00476-w","kind":"doi","is_primary":1}]},{"slug":"efficient-benchmarking-in-production","title":"Efficient Benchmarking in Production","summary":"Empirical study of cheap, recurring evaluation for an evolving production analytics agent, comparing sampling, caching, fixed representative subsets and adaptive testing across 574 historical production benchmark runs.","resource_type":"workflow-process","disciplines":["computer-science"],"tasks":["agent-evaluation"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans set objectives, supply data and check outputs.","source_checked_date":"2026-10-17","limitations":"Author-reported production study; the reported fidelity and cost figures were not independently reproduced. Preprint record only; no venue or artefact release confirmed in checked sources.","adoption_requirements":"Research paper; no code, data or service release confirmed in checked sources. Reuse would require comparable production evaluation logs.","updated_at":"2026-10-17T09:28:23Z","url":"https://research.wusong.net/cases/efficient-benchmarking-in-production","sources":[{"label":"Efficient Benchmarking in Production: A Study of an Evolving LLM Agent - arXiv 2609.21267","url":"https://arxiv.org/abs/2609.21267","kind":"arxiv","is_primary":1}]},{"slug":"electrochromic-thin-film-sdl-process","title":"Self-driving laboratory process for electrochromic thin films","summary":"A self-driving-laboratory process for solution-processed electrochromic thin-film coatings combines automated data acquisition, image processing, spectral analysis and Bayesian optimization to explore processing parameters.","resource_type":"workflow-process","disciplines":["chemistry","materials-science"],"tasks":["data-analysis","image-analysis","optimization"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level workflow automation intentionally unspecified. A reusable component/process reference is not a measured autonomous workflow; researcher oversight remains necessary.","source_checked_date":"2026-10-17","limitations":"Abstract-only substantive review, with additional institutional publication metadata and Wiley-deposited Crossref record. No detailed coating recipe, instrument stack, optimization settings, measured throughput or optical-performance results reviewed or reproduced. Generalization to other solution-processed materials is an author suggestion, not verified. The direct Wiley page was blocked; official Fraunhofer and deposited journal metadata identify the publication.","adoption_requirements":"Editorial, untested: adaptation would require a suitable coating/measurement platform, a defined quality objective, safe materials handling and optimization integration. Exact reagents, parameter bounds, hardware, code and calibration steps are not supplied by the inspected abstract.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/electrochromic-thin-film-sdl-process","sources":[{"label":"A self-driving lab for solution-processed electrochromic thin films","url":"https://arxiv.org/abs/2512.05989","kind":"arxiv","is_primary":1},{"label":"Primary publication/identity evidence for Q112 (112-publication-source.json)","url":"https://publica.fraunhofer.de/entities/publication/7605c78f-1275-4e18-bcd0-8920f0ecb163","kind":"institutional-report","is_primary":1}]},{"slug":"elicit","title":"Elicit","summary":"Elicit offers semantic literature search, cited research reports and assistance with systematic-review screening and data extraction.","resource_type":"tool","disciplines":["multidisciplinary"],"tasks":["data-extraction","knowledge-synthesis","literature-research"],"evidence_level":"vendor-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Homepage feature descriptions only. Marketing accuracy, coverage and time-saving figures are not adopted; search completeness, extraction correctness and citations need independent checking.","adoption_requirements":"An Elicit account/service access, a review question, eligibility criteria and access to original papers.","updated_at":"2026-10-16T14:58:30Z","url":"https://research.wusong.net/cases/elicit","sources":[{"label":"Elicit: AI for research & decision-making","url":"https://elicit.com","kind":"vendor-documentation","is_primary":1}]},{"slug":"emission-gpt","title":"Emission-GPT","summary":"Knowledge-enhanced LLM agent for the atmospheric-emissions domain: domain question answering plus natural-language analysis of emission inventories; case study in Guangdong Province.[20]","resource_type":"tool","disciplines":["computer-science","earth-environmental"],"tasks":["data-analysis","knowledge-synthesis"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans set objectives and validate results.","source_checked_date":"2026-10-16","limitations":"Preprint; not independently reproduced; code availability not verified. Knowledge-base currency not assessed.","adoption_requirements":"Not verified in checked records; knowledge base and LLM access required; assess before adoption.","updated_at":"2026-10-17T08:00:32Z","url":"https://research.wusong.net/cases/emission-gpt","sources":[{"label":"Emission-GPT: A domain-specific language model agent for knowledge retrieval, emission inventory and data analysis - arXiv:2510.02359","url":"https://arxiv.org/abs/2510.02359","kind":"arxiv","is_primary":1}]},{"slug":"environmental-slow-ai-design-principles","title":"Environmental Slow AI: five principles for generative-system design","summary":"A position-paper framework proposing restraint, sufficiency, selectivity over retention, material visibility and friction as affordance as environmental design principles for generative AI.","resource_type":"framework","disciplines":["computer-science","earth-environmental","humanities"],"tasks":["knowledge-synthesis"],"evidence_level":"mixed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level workflow level intentionally unspecified. Conceptual references and reusable process patterns do not themselves establish an autonomous workflow; researchers retain task-specific interpretation and oversight.","source_checked_date":"2026-10-17","limitations":"Abstract-only substantive review plus official ICML workshop-list metadata. Detailed design examples, implementation procedures and environmental measurements were not read. The author's human-agency and sustainability framing is not a verified outcome. The arXiv comment calls this a non-archival conference paper; the official conference list confirms participation but does not establish peer-review status of the specific contribution. Mixed denotes primary abstract and official workshop documentation, not confirmed peer review.","adoption_requirements":"Editorial, untested: use the principles as a discussion lens with environmental and design specialists; establish project-specific environmental measures and rights/privacy constraints before making sustainability claims. No software or installation requirements were established.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/environmental-slow-ai-design-principles","sources":[{"label":"Environmental Slow AI — primary arXiv v1 abstract","url":"https://arxiv.org/abs/2608.20398","kind":"arxiv","is_primary":1},{"label":"ICML 2026 Culture x AI — official workshop contribution list","url":"https://icml.cc/virtual/2026/workshop/54085","kind":"workshop-page","is_primary":1},{"label":"SSRN official record — same-paper identity metadata excerpt","url":"https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7155818","kind":"doi","is_primary":1}]},{"slug":"escriptorium","title":"eScriptorium","summary":"eScriptorium supports segmentation and transcription of historical manuscripts, printed books and archival documents through OCR/HTR and manual correction.","resource_type":"tool","disciplines":["humanities"],"tasks":["archival-research","transcription"],"evidence_level":"project-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"L2 describes the documented human-supervised sequence, not autonomous research. Humans supply inputs, check intermediate results and approve conclusions.","source_checked_date":"2026-10-15","limitations":"Documentation supports the workflow, not a universal recognition accuracy. Script, layout and image quality affect suitability; deployment and model training were not tested.","adoption_requirements":"Document images and reuse rights, access to an eScriptorium deployment, and expertise for transcription correction and ground truth.","updated_at":"2026-10-16T14:58:30Z","url":"https://research.wusong.net/cases/escriptorium","sources":[{"label":"About - eScriptorium","url":"https://escriptorium.eu/about","kind":"project-documentation","is_primary":1}]},{"slug":"evidentia-structured-literature-evidence","title":"Evidentia: human-scoped literature claim extraction and disagreement mapping","summary":"A repository-documented tool for user-uploaded research papers, structured claim extraction, normalised intervention/outcome grouping and fixed study-weighted disagreement views. Engineering demonstrations include synthetic data; neither disagreement scores nor model confidence establish clinical certainty or research validity.","resource_type":"tool","disciplines":["biomedicine","multidisciplinary"],"tasks":["data-analysis","data-extraction","knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"author-reported","workflow_level":"L2","workflow_note":"Scoped L2 for human-selected uploaded papers, text-based claim extraction, grouping and disagreement views. The up-to-eight claim count is a prompt instruction, not an enforced cap; embeddings support organisation, not verified retrieval-augmented extraction. Expert source checking is recommended, not an enforced source-proven gate. Engineering/synthetic outcomes do not establish scientific or clinical validity.","source_checked_date":"2026-10-19","limitations":"LLM extraction is capped at the first12,000characters and prompt-instructed to return up to8claims (an instruction, NOT an enforced output-count limit); parsing/default labels and heuristic fallbacks can conceal missing or incorrect evidence. Without the model key, seeded random vectors and placeholder/heuristic claims are not semantic representations or reliable study findings. Normalised intervention/outcome strings and coarse study-design weights do not establish cross-study comparability, risk of bias or effect magnitude. Model confidence is not calibrated evidence certainty. Demo data and lakehouse checks are explicitly synthetic/engineering-level. No mandatory per-claim expert acceptance/source-span checking, clinical validation, complete literature search or independent reproduction is established.","adoption_requirements":"The documented application uses its Python/API, database/vector and optional model dependencies; cloud/lakehouse components are optional additional infrastructure. Editorial requirements, not verified implementation features: rights/privacy review before sending text to external services; domain expertise and a real-paper error audit; explicit labelling of model/fallback/synthetic output and researcher approval of interpretations.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/evidentia-structured-literature-evidence","sources":[{"label":"Source explains user-uploaded-paper claim extraction, grouping, aggregation and review views.","url":"https://raw.githubusercontent.com/armita-jamshidi/AI_powered_literature_review_engine/9e85e51159f60af541716a1180986a2da9ff6e33/README.md","kind":"code-repository","is_primary":1},{"label":"Source declares the fixed structured-extraction method and prompt.","url":"https://raw.githubusercontent.com/armita-jamshidi/AI_powered_literature_review_engine/9e85e51159f60af541716a1180986a2da9ff6e33/backend/app/pipeline.py","kind":"project-documentation","is_primary":1},{"label":"Lakehouse source attributes seeded-data processing counts to its demo.","url":"https://raw.githubusercontent.com/armita-jamshidi/AI_powered_literature_review_engine/9e85e51159f60af541716a1180986a2da9ff6e33/lakehouse/README.md","kind":"project-documentation","is_primary":1}]},{"slug":"fast-cat-autonomous-catalyst-benchmarking","title":"Fast-Cat: autonomous catalyst benchmarking","summary":"Fast-Cat is a self-driving catalysis laboratory whose authors describe autonomous ligand benchmarking and multi-objective Pareto-front mapping of hydroformylation reactions in a modular flow chemistry platform.","resource_type":"use-case","disciplines":["chemistry","engineering"],"tasks":["experiment-design","experiment-execution","optimization"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":"L3","workflow_note":"Bounded closed-loop experiment selection and execution within a researcher-defined apparatus and chemical task; researchers supply objectives, equipment, initial materials, safety limits and final interpretation. L3 does not mean unrestricted or unsupervised research autonomy.","source_checked_date":"2026-10-17","limitations":"Primary research abstract plus publisher preview only; main methods and supplementary data were not evaluated. No lab execution, numerical performance validation, installation or independent reproduction. High-temperature/high-pressure gas–liquid chemistry needs specialist facilities and safety review. Repository identity checked via API, not code or licence audited.","adoption_requirements":"Specialist chemical laboratory, compatible automation and measurement hardware, trained operators, local risk assessment and institution-approved operating limits; evaluate equipment/software rights and data confidentiality before reuse. Costs, licences and setup completeness were not audited.","updated_at":"2026-10-17T19:24:37Z","url":"https://research.wusong.net/cases/fast-cat-autonomous-catalyst-benchmarking","sources":[{"label":"Autonomous reaction Pareto-front mapping with a self-driving catalysis laboratory","url":"https://www.nature.com/articles/s44286-024-00033-5","kind":"doi","is_primary":1},{"label":"A self-driving lab for accelerated catalyst development","url":"https://doi.org/10.1038/s44286-024-00043-3","kind":"doi","is_primary":0}]},{"slug":"ficsagenticdronesim","title":"FICSAgenticDroneSim","summary":"Repository-documented framework for studying decentralized persistent drone agents with local beliefs in CARLA-Air/AirSim. It documents a mock-simulator path and interchangeable LLM or deterministic planning policies for research on coordination under degraded communication and agent loss.","resource_type":"framework","disciplines":["engineering","robotics"],"tasks":["agent-evaluation","experiment-execution"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation is unspecified. Researchers choose goals, inputs, permissions and evaluation criteria and remain responsible for interpreting outputs.","source_checked_date":"2026-10-17","limitations":"Partial README review: introduction, quick start, interfaces, mission, safety/evaluation caveats and verification sections; no source-code audit or simulation execution. Communication degradation is an experimental model rather than a real-radio model. The README reports unreconciled task boards after partitions; guardian containment is not a real-world safety guarantee. Test/demo totals and comparative effectiveness are not independently verified or adopted as outcomes.","adoption_requirements":"Inspect repository requirements and the mock mission first; CARLA-Air/AirSim and model-service/local-model configuration are needed only for applicable paths. Review data/API privacy and credentials before an LLM-backed run. Keep experiments in simulation; this review grants no real-flight safety assurance. Licence terms require separate inspection.","updated_at":"2026-10-17T20:39:39Z","url":"https://research.wusong.net/cases/ficsagenticdronesim","sources":[{"label":"FICSAgenticDroneSim — README","url":"https://github.com/AkbasLab/FICSAgenticDroneSim","kind":"code-repository","is_primary":1}]},{"slug":"flamepilot","title":"FlamePilot","summary":"FlamePilot combines literature reading with tool-based combustion simulation and error correction. Its authors report executable cases for all 16 FoamBench-Advanced tasks, but a success rate of 0.438 under the benchmark criterion. The combustion case includes researcher-selected changes and manual mesh refinement. Results describe a supervised simulation copilot, not an independently validated autonomous laboratory.","resource_type":"framework","disciplines":["engineering"],"tasks":["research-coding","workflow-orchestration"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":"L3","workflow_note":"L3 applies only to the interactive MILD-combustion simulation workflow. A researcher provides the paper, selects parameter/model changes and manually refines mesh settings; agent code executes the resulting study. Literature extraction, tool execution and log-based repair support researcher-directed simulation. No general researcher-free science claim. Solver executability and agreement with a chosen reference are distinct from physical validation; no universal stopping or safety protocol is certified.","source_checked_date":"2026-10-19","limitations":"Execution and fidelity differ. In the MILD-combustion example, a researcher selects model/parameter changes and manually refines mesh settings; not fully autonomous experimental science.","adoption_requirements":"Untested suggestion: inspect the pinned sources and validate data, evaluator assumptions, implementation availability and researcher controls before planning a bounded trial. No setup or trial was executed.","updated_at":"2026-10-20T08:01:56Z","url":"https://research.wusong.net/cases/flamepilot","sources":[{"label":"Primary manuscript — 2601.01357v1","url":"https://arxiv.org/html/2601.01357v1","kind":"arxiv","is_primary":1},{"label":"Primary project repository or artifact (version limits apply)","url":"https://github.com/deepflame-ai/flamepilot/blob/936012622eec82b3ad6f4e1c74bc9279f948d27d/README.md","kind":"code-repository","is_primary":1}]},{"slug":"foam-agent","title":"Foam-Agent","summary":"Foam-Agent plans, runs and repairs OpenFOAM workflows using solver feedback. In the v3 author evaluation, 97 of 110 Basic cases executed, while only 46 met the stricter execution-and-fidelity criterion. This distinction matters: runnable simulations can still be inaccurate. Public code is available; industrial and experimental validation remain outside these results.","resource_type":"framework","disciplines":["engineering"],"tasks":["research-coding","workflow-orchestration"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":"L3","workflow_note":"L3 applies only to the user-requested OpenFOAM configuration and repair workflow. A user supplies simulation requirements; the author evaluation checks outputs against expert reference fields. Per-iteration human approval is not claimed. Planning generates files; solver logs drive bounded correction. Execution success is separated from field-level fidelity. The loop stops at execution success or its configured iteration cap. Successful execution alone is not physical correctness or unrestricted scientific autonomy.","source_checked_date":"2026-10-19","limitations":"Crash-free CFD is not scientific correctness; tutorial-derived evaluation and solver/model errors limit transfer. Current README or other-model rates are not substituted for v3 results.","adoption_requirements":"Untested suggestion: inspect the pinned sources and validate data, evaluator assumptions, implementation availability and researcher controls before planning a bounded trial. No setup or trial was executed.","updated_at":"2026-10-20T08:01:56Z","url":"https://research.wusong.net/cases/foam-agent","sources":[{"label":"Primary manuscript — 2505.04997v3","url":"https://arxiv.org/html/2505.04997v3","kind":"arxiv","is_primary":1},{"label":"Primary project repository or artifact (version limits apply)","url":"https://github.com/csml-rpi/Foam-Agent/blob/ed8db9415eb7941e47a0a319b611f29e747087f5/README.md","kind":"code-repository","is_primary":1}]},{"slug":"generative-ai-pharmaceutical-lab-monitoring","title":"Generative-AI monitoring of pharmaceutical research laboratories","summary":"Proof-of-concept research application combining generative AI and mobile robotics for monitoring pharmaceutical R&D laboratories.","resource_type":"use-case","disciplines":["biomedicine","robotics"],"tasks":["image-analysis","workflow-orchestration"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"L2 conservatively describes the bounded AI-assisted application in the abstract; no end-to-end autonomous scientific discovery or safety guarantee is asserted. Researchers retain responsibility for validation and safe use.","source_checked_date":"2026-10-17","limitations":"Abstract-only review of arXiv v1. Exact robot integration, model choices, detection accuracy, test protocol and failure handling were not inspected. A proof of concept does not establish regulatory compliance, safety improvement or reliable anomaly detection. No clinical or laboratory deployment is recommended on this evidence alone.","adoption_requirements":"Untested editorial guidance: adaptation requires robotics and laboratory-safety expertise, authorized monitoring data, privacy review and personnel verification of alerts. Determine equipment, model and evaluation requirements from the full methods before prototyping.","updated_at":"2026-10-17T15:13:29Z","url":"https://research.wusong.net/cases/generative-ai-pharmaceutical-lab-monitoring","sources":[{"label":"Autonomous Monitoring of Pharmaceutical R&D Laboratories with 6 Axis Arm Equipped Quadruped Robot and Generative AI: A Preliminary Study — primary arXiv record","url":"https://arxiv.org/abs/2403.10108","kind":"arxiv","is_primary":1}]},{"slug":"genomic-foundation-models-task-first-guide","title":"Genomic foundation models: task-first evidence guide","summary":"Tim Richardson’s guide distinguishes variant calling, post-calling variant-effect scoring and custom DNA-sequence classification, and organises genomic foundation-model choices around the biological task and the kind of supporting evidence.","resource_type":"tutorial-template","disciplines":["biology","biomedicine"],"tasks":["experiment-design","knowledge-synthesis"],"evidence_level":"project-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation is unspecified. Researchers define the task, inspect sources/code and validate outputs; this review did not execute a workflow.","source_checked_date":"2026-10-17","limitations":"Partial-documentation review: forum announcement and guide introduction, decision tree, model overview and opening efficiency discussion were read. The guide is primary evidence of its own tutorial content, not primary validation of the cited models. Downstream papers, benchmark data, cloud costs, code and clinical claims were not audited or reproduced. No model superiority, numeric outcome or clinical validity is endorsed. Not a validated clinical protocol.","adoption_requirements":"Reading requires genomics and model-evaluation literacy. Any subsequent model experiment needs task-appropriate held-out data, checkpoint-specific license review, suitable compute and institutional patient-data governance; these are editorial prerequisites, not a tested recipe.","updated_at":"2026-10-17T17:46:50Z","url":"https://research.wusong.net/cases/genomic-foundation-models-task-first-guide","sources":[{"label":"Genomic foundation models: task-first evidence guide — primary documentation","url":"https://rewirebio.io/blog/genomic-foundation-models-in-the-diagnostic-lab/","kind":"tutorial-blog","is_primary":1},{"label":"Author announcement and ownership disclosure","url":"https://discuss.huggingface.co/t/genomic-foundation-models-an-evidence-guide-for-diagnostic-tasks/190624.json","kind":"community-discussion","is_primary":1}]},{"slug":"gitlab-growth-ai-experiment-skills","title":"GitLab Growth AI skills for experiment implementation","summary":"Repository-documented agent skills for GitLab workflows, including a GLEX A/B/n experiment skill covering implementation, event tracking, rollout and validation. This is experiment-infrastructure guidance, not a statistical analysis method or evidence of improved experimental outcomes.","resource_type":"agent-skill","disciplines":["business","computer-science"],"tasks":["experiment-design","research-coding"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation unspecified; researchers define aims and verify outputs. No autonomous end-to-end outcome claimed.","source_checked_date":"2026-10-17","limitations":"README and partial experiment-instruction review only; other skill implementations and upstream dependency docs not audited. Instructions are GitLab-specific and do not establish randomization quality, statistical validity, or academic effectiveness. Protect private-project and per-person telemetry; retain human runtime review.","adoption_requirements":"OpenCode or a compatible skill-loading coding agent; the experiment skill assumes a GitLab development environment and familiarity with feature flags, Ruby/Vue and GLEX. Installation and access permissions must be reviewed by the adopting team.","updated_at":"2026-10-17T20:06:54Z","url":"https://research.wusong.net/cases/gitlab-growth-ai-experiment-skills","sources":[{"label":"GitLab.org / Growth / ai / skills","url":"https://gitlab.com/gitlab-org/growth/ai/skills/-/raw/main/README.md","kind":"code-repository","is_primary":1},{"label":"GitLab Experiment skill instructions","url":"https://gitlab.com/gitlab-org/growth/ai/skills/-/raw/main/skills/gitlab-experiment/SKILL.md","kind":"project-documentation","is_primary":1}]},{"slug":"guideline-constrained-llm-checklist-acceptability","title":"Guideline-constrained LLM checklist generation and clinician acceptability evaluation","summary":"A study uses ChatGPT 5.2 to convert an adult neck-mass guideline into a history-and-examination checklist, then evaluates clinician perceptions through a survey and thematic analysis of free-text feedback. The authors report 22 respondents and incomplete translation of acceptability into willingness to adopt.","resource_type":"workflow-process","disciplines":["medicine"],"tasks":["data-extraction","knowledge-synthesis"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Researcher-supervised process; automation level intentionally unspecified. No autonomous research or clinical deployment asserted.","source_checked_date":"2026-10-17","limitations":"Publisher paper-excerpt review via extracted text; figures and source guideline were not independently audited. One guideline, one model and a modest clinician sample; perceived utility is not real-world use or patient outcomes. The authors report missing airway-related symptoms, family history and scope examination. No independent replication or safety certification. The safety-item table records one disagree response, despite discussion text saying no response suggested unsafe. This discrepancy is not reconciled here; clinician perceptions are not proof of safety.","adoption_requirements":"Access rights to the guideline and a suitable LLM; qualified clinician oversight, documented omissions checks, and applicable ethics/privacy approvals for participant research. No patient data upload required or authorized by this review.","updated_at":"2026-10-17T18:33:50Z","url":"https://research.wusong.net/cases/guideline-constrained-llm-checklist-acceptability","sources":[{"label":"Acceptability of a Large Language Model (LLM)-Generated Guideline-Based Checklist Among Otolaryngologists: A Cross-Sectional Survey and Thematic Analysis","url":"https://doi.org/10.7759/cureus.106804","kind":"doi","is_primary":1}]},{"slug":"handian-chongguang","title":"汉典重光 (Handian Chongguang)","summary":"Handian Chongguang is a documented AI-assisted digitisation and research platform for historical Chinese books, announced by Sichuan University with Alibaba and UC Berkeley partners.","resource_type":"use-case","disciplines":["humanities"],"tasks":["archival-research","transcription"],"evidence_level":"institutional-report","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"L2 describes the documented human-supervised sequence, not autonomous research. Humans supply inputs, check intermediate results and approve conclusions.","source_checked_date":"2026-10-15","limitations":"Historical launch report, not a current platform audit. Recognition accuracy claims are not independently validated or repeated here. Present access, corpus availability, rights and implementation details need checking before adoption.","adoption_requirements":"Access to relevant historical-book images or the platform if still available, Chinese palaeographic expertise and reuse permissions.","updated_at":"2026-10-16T14:58:30Z","url":"https://research.wusong.net/cases/handian-chongguang","sources":[{"label":"川大携手阿里、伯克利发布“汉典重光” 中国海外古籍“数字化回归”平台","url":"https://news.scu.edu.cn/info/1153/36677.htm","kind":"institutional-report","is_primary":1}]},{"slug":"haowjy-autoresearch-meridian","title":"autoresearch — Meridian research-loop prompts","summary":"Repository-documented prompt package for coordinating AI research loops in the Mars/Meridian ecosystem, with research-tree exploration, investigation design, run supervision and preserved research memory.","resource_type":"agent-skill","disciplines":["multidisciplinary"],"tasks":["experiment-design","workflow-orchestration"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation unspecified. Researchers approve objectives and inspect outputs; documentation is not an executed workflow.","source_checked_date":"2026-10-17","limitations":"README-only source review by an AI editorial agent; no implementation execution, independent outcome evaluation, security audit or full linked-document review. HEAD is mutable; saved content and hashes preserve the inspected version. Same-name autoresearch projects are not presumed identical. GitHub identifies this as haowjy/autoresearch, not a fork; Mars/Meridian are dependencies/ecosystem, not identity aliases.","adoption_requirements":"Mars/Meridian downstream project and the named meridian-dev-workflow/design-lead dependencies; exact installation and compatibility have not been audited.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/haowjy-autoresearch-meridian","sources":[{"label":"haowjy/autoresearch — README","url":"https://raw.githubusercontent.com/haowjy/autoresearch/HEAD/README.md","kind":"code-repository","is_primary":1}]},{"slug":"harness-engineering-coding-agent-configuration","title":"Harness engineering: repository-level coding-agent configuration","summary":"An exploratory study of repository-level configuration for coding agents, including context files, skills and subagents. The authors report analysing 2,853 GitHub repositories across five coding tools.","resource_type":"workflow-process","disciplines":["computer-science"],"tasks":["research-coding","workflow-orchestration"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Resource-level automation unspecified; researchers select the task, assess data rights and inspect outputs. Source review is not an executed workflow.","source_checked_date":"2026-10-17","limitations":"Abstract-only review of arXiv v5. The record says it is based on the earlier conference paper Configuring Agentic AI Coding Tools: An Exploratory Study; this is a title/version relationship, not a separately verified publication. Repository sampling, classification and data artifacts were not inspected. Counts and adoption findings are author-reported; no performance gain or independent reproduction is established.","adoption_requirements":"Editorial, untested: review the full methods, check current tool-specific configuration formats, and version project instructions. Do not put secrets or restricted research data in agent context; permissions and output testing remain the researcher's responsibility.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/harness-engineering-coding-agent-configuration","sources":[{"label":"Harness Engineering for Agentic AI Coding Tools: An Exploratory Study","url":"https://arxiv.org/abs/2602.14690","kind":"arxiv","is_primary":1}]},{"slug":"harnessnet","title":"HarnessNet","summary":"Framework for hierarchical joint harness evolution in multi-agent systems, with Agentron units forming layered graphs and backward feedback propagation; evaluated on AgentBench.[19]","resource_type":"framework","disciplines":["computer-science"],"tasks":["agent-evaluation","workflow-orchestration"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans set objectives and validate results.","source_checked_date":"2026-10-16","limitations":"Preprint (2026); not independently reproduced; code availability not verified.","adoption_requirements":"Multi-agent system harness code and AgentBench-style evaluation environment; no packaged implementation verified.","updated_at":"2026-10-17T08:00:32Z","url":"https://research.wusong.net/cases/harnessnet","sources":[{"label":"HarnessNet: Feedback-driven Hierarchical Joint Harness Evolution for Multi-agent Systems - arXiv:2610.15221","url":"https://arxiv.org/abs/2610.15221","kind":"arxiv","is_primary":1}]},{"slug":"hierarchical-deep-research-local-web-rag","title":"Hierarchical Deep Research with Local-Web RAG: Toward Automated System-Level Materials Discovery","summary":"A hierarchical deep-research framework for materials and device discovery combines local retrieval-augmented generation with language-model reasoning and a Deep Tree of Research that expands and prunes research branches.","resource_type":"framework","disciplines":["materials-science"],"tasks":["literature-research","materials-discovery"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Automation level unspecified: a resource-level description is not a measured autonomous workflow. Researchers select goals and data and validate outputs.","source_checked_date":"2026-10-17","limitations":"Primary arXiv abstract-only review; full text and software not reviewed or executed. Judge-based report assessment and author-reported dry-lab checks do not establish experimentally realized materials or independent validation. Code availability, hardware and exact costs were not verified.","adoption_requirements":"Editorial, untested guidance: assess access to local literature, a language-model reasoner and domain simulation expertise; do not assume a turnkey distribution. Review rights and confidentiality before indexing local documents.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/hierarchical-deep-research-local-web-rag","sources":[{"label":"Hierarchical Deep Research with Local-Web RAG: Toward Automated System-Level Materials Discovery — arXiv 2511.18303","url":"https://arxiv.org/abs/2511.18303","kind":"arxiv","is_primary":1}]},{"slug":"hitl-research-agent-source-analysis","title":"HITL Research Agent: source analysis and planned research workflow","summary":"An in-development human-in-the-loop research framework. Its README reports a working single-article analysis step producing structured JSON and PDF preparation into text and crops. Cross-source synthesis, literature discovery, retrieval and a separate evaluation stage are designed but not yet built; researchers remain responsible for source checking and acceptance.","resource_type":"framework","disciplines":["multidisciplinary"],"tasks":["data-extraction","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation unspecified; researchers define aims and verify outputs. No autonomous end-to-end outcome claimed.","source_checked_date":"2026-10-17","limitations":"README-only substantive review plus repository identity metadata. No installation, API call, pilot execution or evaluation reproduced. author_stated is a model attribution, not verification; the current step does not retrieve cited studies. Planned A/B experiments are not reported results.","adoption_requirements":"README specifies Python 3.12, uv, git 2.31+ and an OpenAI API key. Use only articles whose rights and confidentiality permit transmission to the model provider; budget for model calls and researcher review.","updated_at":"2026-10-17T20:06:54Z","url":"https://research.wusong.net/cases/hitl-research-agent-source-analysis","sources":[{"label":"kaseymallette/hitl-research-agent","url":"https://raw.githubusercontent.com/kaseymallette/hitl-research-agent/main/README.md","kind":"code-repository","is_primary":1}]},{"slug":"hler","title":"HLER","summary":"HLER connects dataset-aware economic questions, human selection gates, econometric execution and review-driven revision. Authors report 69 of 79 dataset-aware questions were feasible, compared with 34 of 82 unconstrained questions, and 12 of 14 runs completed. Automated reviewer gains are same-model assessments. The public repository contains working papers and per-study scripts, explicitly excluding the research system; causal validity and independent replication remain unverified.","resource_type":"workflow-process","disciplines":["economics"],"tasks":["data-analysis","hypothesis-generation"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":"L3","workflow_note":"L3 applies only to the dataset-constrained economic-research pipeline. The human principal investigator chooses the question and retains publication approval; dataset audit and author evaluation bound interpretation. Agents execute econometrics and review-driven revision. The public output repository excludes the agent system. Runs can fail to converge; same-model review is not independent causal validation, and multiple-testing/identification limits remain.","source_checked_date":"2026-10-19","limitations":"Observational identification, multiple testing, limited estimator support and small sample remain. APIcost/runtime are historic author reports, not reproduced/current quotes. Manuscript outputs do not prove causal discoveries. Body-date/arXiv-date inconsistency remains unresolved.","adoption_requirements":"Untested suggestion: inspect the pinned sources and validate data, evaluator assumptions, implementation availability and researcher controls before planning a bounded trial. No setup or trial was executed.","updated_at":"2026-10-20T08:01:56Z","url":"https://research.wusong.net/cases/hler","sources":[{"label":"Primary manuscript — arXiv2603.07444v1;8March2026;body also prints24August2026","url":"https://arxiv.org/html/2603.07444v1","kind":"arxiv","is_primary":1},{"label":"Primary project repository or artifact (version limits apply)","url":"https://github.com/maxwell2732/hler-working-papers/blob/0bfd0eeecfa848c82eb7b6f6cc04fb182f76299c/README.md","kind":"code-repository","is_primary":1}]},{"slug":"human-aware-robot-behaviour-sdl","title":"Human-aware robot behaviour for shared laboratory equipment","summary":"An initial AI-driven perception and human-intention prediction approach for mobile robot chemists sharing laboratory instruments with researchers.","resource_type":"framework","disciplines":["chemistry","robotics"],"tasks":["experiment-execution","workflow-orchestration"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level workflow automation intentionally unspecified. A reusable component/process reference is not a measured autonomous workflow; researcher oversight remains necessary.","source_checked_date":"2026-10-17","limitations":"Abstract-only for the promoted method, plus a selected ACM conclusion excerpt and official publication/program/policy metadata. The ACM extraction was additionally inspected and cautions that broader implementation and workflow-efficiency evaluation remain future work; no quantitative accuracy is promoted. No controller execution, safety validation, user study or experimental reproduction was performed. Peer-reviewed means a published HRI Late-Breaking Report, not a full-length HRI paper or independently validated deployment.","adoption_requirements":"Editorial, untested: adaptation would require compatible mobile-robot sensing/control and laboratory orchestration, explicit human-interaction procedures and independent safety assessment. Do not treat predicted intent as consent or as a replacement for physical safety controls. Exact hardware and model setup are not prescribed by the inspected abstract.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/human-aware-robot-behaviour-sdl","sources":[{"label":"Human-Aware Robot Behaviour in Self-Driving Labs","url":"https://arxiv.org/abs/2603.08420","kind":"arxiv","is_primary":1},{"label":"Primary publication/identity evidence for Q113 (113-publication-source.json)","url":"https://livrepository.liverpool.ac.uk/3203745","kind":"institutional-report","is_primary":1}]},{"slug":"human-rights-based-responsible-ai","title":"Human-rights-based responsible AI: a harm-centred research lens","summary":"A normative framework proposing universal human rights as an explicit value-alignment basis for responsible AI research and civil-society participation, centring who is harmed, which harms they face and possible mitigation.","resource_type":"framework","disciplines":["computer-science","law","social-science"],"tasks":["knowledge-synthesis"],"evidence_level":"mixed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level workflow level intentionally unspecified. Conceptual references and reusable process patterns do not themselves establish an autonomous workflow; researchers retain task-specific interpretation and oversight.","source_checked_date":"2026-10-17","limitations":"ArXiv abstract-only substantive review plus an extracted official conference poster; paper body and visual poster elements were not reviewed. The non-archival EAAMO poster is verified, but archival peer-reviewed publication is not established. Mixed denotes abstract and conference-poster evidence, not a verified peer-review judgment. Universality, cultural relativism and legal enforceability remain issues, not solved properties. This is not legal advice, an operational compliance system or a validated harm-reduction intervention.","adoption_requirements":"Editorial, untested: identify affected communities, relevant rights and local legal context with appropriate domain and civil-society expertise. Participation, legal interpretation and harm measurement need project-specific work; no installation or automatic compliance capability is asserted.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/human-rights-based-responsible-ai","sources":[{"label":"Human Rights-Based Approach — arXiv v1 abstract","url":"https://arxiv.org/abs/2210.02667","kind":"arxiv","is_primary":1},{"label":"EAAMO 2022 official conference poster — extracted text","url":"https://conference2022.eaamo.org/posters/121_prabhakaran.pdf","kind":"workshop-page","is_primary":1}]},{"slug":"humanities-ai-research-infrastructure","title":"Humanities AI Research Infrastructure","summary":"Scholar-documented humanities research stack for concept genealogy, literature-grounded argument development and media/platform analysis.","resource_type":"workflow-process","disciplines":["arts","humanities"],"tasks":["knowledge-synthesis","literature-research","scientific-writing"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Workflow level unspecified for this resource; researchers define objectives, control input data and validate outputs. Suggested adoption steps were not exercised.","source_checked_date":"2026-10-17","limitations":"README-only review; directory contents, installation and session outputs not independently validated. Author says infrastructure is under active construction and not a beginner tutorial. The theoretical framework and claimed model/corpus configuration are author descriptions, not empirical validation; interpretive claims and citations require scholarly checking.","adoption_requirements":"Technical setup for the documented Khoj/Qdrant and DeerFlow stack, own lawful research corpus and model-provider access; README warns this is not an absolute-beginner tutorial.","updated_at":"2026-10-17T11:14:58Z","url":"https://research.wusong.net/cases/humanities-ai-research-infrastructure","sources":[{"label":"icerain-cmd/humanities-ai-research","url":"https://github.com/icerain-cmd/humanities-ai-research","kind":"code-repository","is_primary":1},{"label":"Humanities AI Research README","url":"https://raw.githubusercontent.com/icerain-cmd/humanities-ai-research/main/README.md","kind":"project-documentation","is_primary":1}]},{"slug":"iflytek-research-model","title":"讯飞科研大模型 (iFLYTEK research model)","summary":"The iFLYTEK research model is offered for paper reading, review generation, specialist translation and English-language polishing.","resource_type":"tool","disciplines":["humanities","multidisciplinary","social-science"],"tasks":["knowledge-synthesis","literature-research","scientific-writing"],"evidence_level":"vendor-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Product claims are vendor-described, not an evaluated research outcome. Translation, summaries and answers require source checking; pricing, access eligibility, privacy and uptime were not independently verified.","adoption_requirements":"Appropriate service/API access, permitted research documents and reviewers able to check scientific and translated content.","updated_at":"2026-10-16T14:58:30Z","url":"https://research.wusong.net/cases/iflytek-research-model","sources":[{"label":"科研大模型_讯飞星火_科技文献语料训练","url":"https://www.xfyun.cn/services/kjwx","kind":"vendor-documentation","is_primary":1}]},{"slug":"interdisciplinary-llm-research-task-reference","title":"LLM research planning: interdisciplinary task examples and constraints","summary":"A conceptual research-planning reference discussing LLM uses in literature summarization, code syntax correction and scientific writing alongside dataset bias and ethical constraints.","resource_type":"tutorial-template","disciplines":["multidisciplinary"],"tasks":["literature-research","research-coding","scientific-writing"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation classification unspecified. This is a research/reference listing, not an exercised workflow; researchers select objectives and check interpretation and outputs.","source_checked_date":"2026-10-17","limitations":"Abstract-only substantive review. The conceptual discussion supplies task examples and constraints but no inspected implementation recipe, validated checklist, comparative experiment or transferable effect estimate. Reported examples are discussion topics, not independently established capabilities or measured acceleration. No peer-reviewed venue was confirmed in the retrieved arXiv record or title search; the listing retains preprint status. It is not a comprehensive or current survey of all LLM research use.","adoption_requirements":"Editorial, untested use: treat the listed tasks and constraints as discussion prompts. Check domain-specific literature, data rights, privacy and output validity before choosing any particular model or research workflow.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/interdisciplinary-llm-research-task-reference","sources":[{"label":"An Interdisciplinary Outlook — primary arXiv abstract","url":"https://arxiv.org/abs/2311.04929","kind":"arxiv","is_primary":1}]},{"slug":"investmentresearchai-probabilistic-model-method","title":"InvestmentResearchAI: probabilistic modelling and evaluation protocol","summary":"Public research documentation describes probabilistic asset-relative modelling, time-separated evaluation and explicit limitations, with a scheduled artifact-generation definition.","resource_type":"use-case","disciplines":["business","computer-science","economics"],"tasks":["data-analysis","experiment-design"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Intended L2 for researcher-configured economic modelling: time-ordered data and splits, a small MLP ensemble, calibration, shadow evaluation, challenger criteria and prospective records. A separate news-labelling protocol prescribes blinded labels and held-out comparison. Scheduled automation is not evidence of successful runs, agency or predictive advantage.","source_checked_date":"2026-10-19","limitations":"README, model method, news-labelling guide and one workflow definition were read as text. Scripts, weights, generated artefacts, news labels, accounts and the deployed UI were not accessed or executed. CI schedules and permissions are not successful-run evidence. Demo records may remain until valid results exist. The source discloses missing point-in-time universe data, economic cost/performance evidence, formal PBO/DSR and adaptive online conformal evaluation. Cited studies were not audited. No profitability, financial advice, trading action or validated research result is established. Moving HEAD documents are unpinned and non-atomic.","adoption_requirements":"Untested suggestion: establish legitimate data, point-in-time lineage, independent evaluation and access boundaries before a research trial.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/investmentresearchai-probabilistic-model-method","sources":[{"label":"Project README","url":"https://raw.githubusercontent.com/Apepsis/InvestmentResearchAI/HEAD/README.md","kind":"code-repository","is_primary":1},{"label":"Model research protocol","url":"https://raw.githubusercontent.com/Apepsis/InvestmentResearchAI/HEAD/NEURAL_MODEL_V8.md","kind":"project-documentation","is_primary":1},{"label":"News-labelling guide","url":"https://raw.githubusercontent.com/Apepsis/InvestmentResearchAI/HEAD/NEWS_LABELING_GUIDE.md","kind":"project-documentation","is_primary":1},{"label":"Workflow definition (documentation only)","url":"https://raw.githubusercontent.com/Apepsis/InvestmentResearchAI/HEAD/.github/workflows/update-market-data.yml","kind":"project-documentation","is_primary":1}]},{"slug":"invisible-hands-for-economists","title":"Invisible Hands for Economists","summary":"Invisible Hands for Economists documents an economics agent skill and artifact-audit harness for empirical, theoretical and structural research. It prescribes source acquisition, frozen research contracts, provenance/result bindings and substantive review, but the deterministic runner does not itself create scholarly evidence.","resource_type":"agent-skill","disciplines":["economics"],"tasks":["data-analysis","literature-research","scientific-writing","workflow-orchestration"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L4","workflow_note":"Protocol-only L4: an explicitly budgeted host-agent research loop with researcher scope/design controls—not the artifact checker and not demonstrated autonomous science. The skill prescribes a fixed baseline, falsifiable hypothesis, attributable change, validation and a recorded keep/discard/inconclusive/blocked/crash decision, bounded by goals, budget and stop conditions. Without an explicit autonomy budget, only baseline, feasibility and planning are prescribed. Research agents must perform the substantive work; the deterministic runner audits supplied artifacts. Source-prescribed researcher checkpoints and data/safety constraints are not evidence that a real human-run trace was observed; named reviewers may be AI. Synthetic engineering/build tests and blocked manuscript examples do not establish scientific novelty, causal validity, empirical success or automatic paper completion. This is a documentation-supported protocol classification, with no source execution or independently reproduced outcomes.","source_checked_date":"2026-10-17","limitations":"Documentation-only review; no repository script, MCP server, research workflow, analysis, proof checker or experiment was executed. The implementation-status document reports synthetic/illustrative engineering cases, with all three manuscript packages blocked; these do not establish real empirical research, valid causal identification, completed publication-quality manuscripts or a one-shot success rate. Runner checks declared artifacts and hashes, not the correctness of economic interpretation or the quality/independence of agent review. Native agent dispatch and scientific work remain host responsibilities. Working-copy replication is explicitly not an OS sandbox; absolute paths can escape it. The status document also records no cross-process locking and wall-clock rather than API-money budget accounting. Token-saving or policy-learning benefits are not verified. Restricted-data approval, provider behavior and dependency terms must be assessed per deployment; the protocol is not proof of privacy enforcement.","adoption_requirements":"Needs a compatible agent host able to perform actual source acquisition and native task dispatch. SKILL.md lists Python 3.10+ with pandas/NumPy for profiling and Lean 4/Mathlib for formal-theory verification; README specifies Python 3.11+ and an optional MCP dependency for that server. End-to-end manuscript work additionally depends on permitted data, analysis/LaTeX tooling and substantive review. Approve command/filesystem/network scope first: the documented replication copy is not an OS sandbox. Follow DUA/IRB and approved model-endpoint rules; no confidential rows, logs, embeddings or small cells may leave the authorized boundary merely because a tool is local.","updated_at":"2026-10-17T22:18:09Z","url":"https://research.wusong.net/cases/invisible-hands-for-economists","sources":[{"label":"Invisible Hands for Economists — primary repository README","url":"https://github.com/Tohskcid/invisible-hands-for-economists","kind":"code-repository","is_primary":1},{"label":"Invisible Hands for Economists — econ-research-lab SKILL.md (read as data)","url":"https://raw.githubusercontent.com/Tohskcid/invisible-hands-for-economists/HEAD/SKILL.md","kind":"project-documentation","is_primary":1},{"label":"Invisible Hands for Economists — implementation and engineering status","url":"https://raw.githubusercontent.com/Tohskcid/invisible-hands-for-economists/HEAD/docs/research/workflow-implementation-status.md","kind":"project-documentation","is_primary":1},{"label":"Invisible Hands for Economists — research execution guide","url":"https://raw.githubusercontent.com/Tohskcid/invisible-hands-for-economists/HEAD/references/workflow_execution.md","kind":"project-documentation","is_primary":1},{"label":"Invisible Hands for Economists — restricted-data boundary protocol","url":"https://raw.githubusercontent.com/Tohskcid/invisible-hands-for-economists/HEAD/references/confidential_data.md","kind":"project-documentation","is_primary":1},{"label":"Invisible Hands for Economists — empirical inference protocol","url":"https://raw.githubusercontent.com/Tohskcid/invisible-hands-for-economists/HEAD/references/empirical.md","kind":"project-documentation","is_primary":1},{"label":"Invisible Hands for Economists — repository MIT LICENSE","url":"https://raw.githubusercontent.com/Tohskcid/invisible-hands-for-economists/HEAD/LICENSE","kind":"project-documentation","is_primary":1}]},{"slug":"isabel-quiceno-behavioral-vulnerability-map","title":"AI Behavioral Research: behavioral vulnerability map","summary":"Spanish-language behavioral evaluation map in Isabel Quiceno’s research portfolio. It organizes LLM failures into psychological manipulation, training biases, persuasion/information integrity and alignment loss, pairing observable behavior with metrics and mitigations. Proposed metrics are explicitly unvalidated.","resource_type":"framework","disciplines":["computer-science","psychology"],"tasks":["agent-evaluation","experiment-design"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation is unspecified. Researchers choose questions, evaluation design and decisions; documentation review is not workflow execution.","source_checked_date":"2026-10-17","limitations":"Repository README and vulnerability map reviewed, not the entire portfolio or its cited literature. No model evaluation reproduced. Red-teaming report and AI UX benchmark are marked in preparation in README; they are not completed empirical results. Proposed metrics and mitigations must not be presented as validated interventions.","adoption_requirements":"Read Spanish-language documentation, distinguish established-source descriptions from the author’s proposed measures, and design simulated scenarios with qualified researcher oversight. Do not use real participant data without ethics/privacy safeguards. License not audited.","updated_at":"2026-10-17T18:33:50Z","url":"https://research.wusong.net/cases/isabel-quiceno-behavioral-vulnerability-map","sources":[{"label":"isabelusuga/ai-behavioral-research-portfolio","url":"https://raw.githubusercontent.com/isabelusuga/ai-behavioral-research-portfolio/main/02-mapa-vulnerabilidades/mapa.md","kind":"code-repository","is_primary":1}]},{"slug":"jahura7-stroke-risk-category-classification","title":"Stroke-risk category classification — educational ML workflow","summary":"An educational machine-learning workflow for classifying a supplied dataset’s Low, Moderate and High stroke-risk labels. Repository documentation describes exploratory analysis, preprocessing, feature engineering and comparison of a linear-regression baseline with a random-forest classifier.","resource_type":"tutorial-template","disciplines":["biomedicine","education"],"tasks":["data-analysis","research-coding"],"evidence_level":"repository-documented","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Resource-level automation is unspecified; researchers select tasks, review sources, validate outputs and authorize consequential actions.","source_checked_date":"2026-10-17","limitations":"README-only review; no notebook execution, dataset inspection or clinical validation. Dataset provenance and label generation are not independently established. The README cautions that linear regression is not designed for categorical classification; nominal encoding and train-only preprocessing require audit, not an inference of executed defects. README notebook/data filenames are inconsistent across sections.","adoption_requirements":"Python/Colab and pandas, NumPy, plotting packages and scikit-learn; inspect the experiment branch and actual filenames before running. Kaggle data retain separate terms. Use only for education; no diagnosis or treatment decisions and no private patient uploads.","updated_at":"2026-10-17T20:06:54Z","url":"https://research.wusong.net/cases/jahura7-stroke-risk-category-classification","sources":[{"label":"Stroke-risk category classification — educational ML workflow","url":"https://github.com/Jahura7/assignment_on_research_design_experiment_and_writing_reseaech_paper","kind":"code-repository","is_primary":1}]},{"slug":"juan-daniel-research-agent-notebook","title":"Research agent: local research notebook and literature brief","summary":"Repository-documented local research notebook with a literature-screening workflow, dated AI logs and a researcher decision inbox. The README describes title, abstract and full-paper screening and a compact reading selection; it does not demonstrate scientific discovery.","resource_type":"tool","disciplines":["multidisciplinary","physics"],"tasks":["literature-research","workflow-orchestration"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation is unspecified. Researchers choose questions, evaluation design and decisions; documentation review is not workflow execution.","source_checked_date":"2026-10-17","limitations":"README-only documentation review; no code execution or scientific outcomes reproduced. The experimental-qubit vault, people, papers and numerical results are explicitly fictional/synthetic. Relevance scores are reading priorities, not scientific quality. License and maintenance not audited.","adoption_requirements":"README requires uv, private configuration and a signed-in Codex CLI; model calls may incur cost. Keep real notes, author mappings, credentials and backups private. Obsidian is optional; task queries require its Tasks plugin.","updated_at":"2026-10-17T18:33:50Z","url":"https://research.wusong.net/cases/juan-daniel-research-agent-notebook","sources":[{"label":"Juan Daniel Torres / AI-Research-Assistant","url":"https://gitlab.com/juandaanieel/ai-research-assistant/-/raw/main/README.md","kind":"code-repository","is_primary":1}]},{"slug":"jupyter-ai","title":"Jupyter AI","summary":"Jupyter AI integrates agent chats into JupyterLab so researchers can collaborate with selected agents in their notebook workspace.","resource_type":"tool","disciplines":["multidisciplinary"],"tasks":["data-analysis","research-coding"],"evidence_level":"project-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"The saved documentation describes the then-current agent-based interface, not older magic-command interfaces. No agent is bundled by default; agent authentication, adapters, permissions and generated code require separate checks.","adoption_requirements":"JupyterLab/Python environment plus a separately installed and authenticated supported agent and any required adapter.","updated_at":"2026-10-16T14:58:30Z","url":"https://research.wusong.net/cases/jupyter-ai","sources":[{"label":"Getting Started - Jupyter AI","url":"https://jupyter-ai.readthedocs.io/en/stable/getting-started.html","kind":"project-documentation","is_primary":1}]},{"slug":"k-agents-quantum-laboratory","title":"k-agents: knowledge-based agents for quantum laboratory experiments","summary":"k-agents uses LLM-based agents to organize laboratory knowledge and automate multi-step experimental procedures, demonstrated by the authors on a superconducting quantum processor.","resource_type":"framework","disciplines":["computer-science","physics"],"tasks":["experiment-execution","workflow-orchestration"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Workflow level intentionally unspecified for a reusable component or conceptual reference; researcher oversight and task-specific validation remain necessary.","source_checked_date":"2026-10-17","limitations":"ArXiv v2 abstract reviewed for the promoted framework description; independent AI audit additionally checked official Patterns publication metadata and selected excerpts, not the complete article. Agent prompts, integration setup, model access, runtime costs and safety controls were not inspected as an executable recipe. Calibration and entangled-state results are author-reported, not reproduced.","adoption_requirements":"The described application involves laboratory-operation interfaces, analysis methods and a superconducting quantum processor. Exact integration dependencies and model costs were not established from the abstract.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/k-agents-quantum-laboratory","sources":[{"label":"Agents for self-driving laboratories applied to quantum computing","url":"https://arxiv.org/abs/2412.07978","kind":"arxiv","is_primary":1},{"label":"Official Patterns publication metadata and selected excerpts","url":"https://doi.org/10.1016/j.patter.2025.101372","kind":"doi","is_primary":1}]},{"slug":"kalaris-research-agent-skills","title":"Research Agent Skills (Kalaris Labs): systematic-review guidance","summary":"A research-skills collection with a documented systematic-review guide linking protocol, search, deduplication, screening, appraisal, synthesis and PRISMA reporting.","resource_type":"agent-skill","disciplines":["multidisciplinary"],"tasks":["knowledge-synthesis","literature-research","scientific-writing"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Intended L2 for the selected systematic-review guide: protocol, search records, deduplication, independent screening, appraisal, appropriate synthesis and reporting. AI-assisted screening must be validated and never the sole reviewer. No universal level for the collection, independent invention of adapted skills, verified script enforcement or research-gain claim.","source_checked_date":"2026-10-19","limitations":"Selected README introduction, research tasks, benchmark context, security/FAQ/citation/licence sections plus full systematic-review guide and third-party notices read; not all collection entries/code/benchmarks audited. Notices list159 adapted K-Dense and98 Orchestra skills; counts are source claims and adaptations do not establish official upstream affiliation or separate research gains. Small author benchmark/pilot claims are not independently adjudicated and no performance/savings figures are promoted. Security/signature/test claims are unverified. Tools run with user permissions. L2 applies to the selected guide, not all280+ skills. AI editorial source review only; no installation, source code execution, provider/model calls or outcome reproduction.","adoption_requirements":"A compatible agent harness and relevant database/data rights; inspect selected instructions/scripts and provider/data handling before use. Installer alternatives and future-release limitations are documentation statements; none were run.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/kalaris-research-agent-skills","sources":[{"label":"KalarisLabs/research-agent-skills primary-readme","url":"https://raw.githubusercontent.com/KalarisLabs/research-agent-skills/df088027ff233dc511fa74cb7f406b4e87fa438c/README.md","kind":"code-repository","is_primary":1},{"label":"KalarisLabs/research-agent-skills systematic-review-guide","url":"https://raw.githubusercontent.com/KalarisLabs/research-agent-skills/df088027ff233dc511fa74cb7f406b4e87fa438c/docs/guides/systematic-review.md","kind":"project-documentation","is_primary":1},{"label":"KalarisLabs/research-agent-skills third-party-notices","url":"https://raw.githubusercontent.com/KalarisLabs/research-agent-skills/df088027ff233dc511fa74cb7f406b4e87fa438c/THIRD_PARTY_NOTICES.md","kind":"project-documentation","is_primary":1}]},{"slug":"khmer-short-answer-grading-research","title":"Khmer short-answer grading: human-orchestrated modelling and SHAP research","summary":"An undergraduate research codebase and documentation for comparing Khmer short-answer grading models and inspecting SHAP word attribution. Its prototype supports research and teacher assistance; reported agreement is with one teacher, not independently established correctness or classroom benefit.","resource_type":"use-case","disciplines":["computer-science","education","linguistics"],"tasks":["data-analysis","experiment-design","experiment-execution"],"evidence_level":"repository-documented","outcome_verification":"author-reported","workflow_level":"L2","workflow_note":"Scoped L2 for human-designed teacher-labelled data, modelling/comparison, evaluation and SHAP/prototype research. Researchers choose stages and comparisons. Single-grader targets, unequal variants/samples, historical provenance and consent gaps prevent high-stakes grading or classroom-effectiveness claims.","source_checked_date":"2026-10-18","limitations":"README and primary documentation reviewed, not full report-PDF review or a rerun. Single-grader labels, student-derived references, unequal comparison/SHAP samples, unresolved prediction/checkpoint/preprocessing provenance and lack of verified consent/ethics evidence restrict claims. Later dataset correction did not regenerate report benchmarks. Public pseudonymous copies are not guaranteed anonymous or authorised for reuse.","adoption_requirements":"Authorised data access and institutional privacy/consent review before handling examination answers; appropriate researcher/teacher expertise, comparable dataset/split/model provenance, suitable environments and checkpoints for any later reproduction. Separate source-code, dataset and model terms apply. Do not use as an autonomous high-stakes grader.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/khmer-short-answer-grading-research","sources":[{"label":"README describes teacher-labelled Khmer modelling/comparison, SHAP and prototype","url":"https://github.com/PhorkNorak/Explainable-AI-for-Automated-Grading-of-Khmer-Language-Short-Answer-Questions-in-Education","kind":"code-repository","is_primary":1},{"label":"Evidence notes disclose differing variants, unresolved historical prediction/provenance audits, SHAP limits and no benchmark rerun.","url":"https://raw.githubusercontent.com/PhorkNorak/Explainable-AI-for-Automated-Grading-of-Khmer-Language-Short-Answer-Questions-in-Education/90f127cc160b3e73aacb9b7c2977295332c9aac6/docs/guides/report_evidence.md","kind":"project-documentation","is_primary":1},{"label":"Ethics notes distinguish public pseudonymous copies from consent/governance evidence and recommend human-reviewed assistance, not high-stakes autonomy.","url":"https://raw.githubusercontent.com/PhorkNorak/Explainable-AI-for-Automated-Grading-of-Khmer-Language-Short-Answer-Questions-in-Education/90f127cc160b3e73aacb9b7c2977295332c9aac6/docs/guides/ethics.md","kind":"project-documentation","is_primary":1},{"label":"Report-folder documentation makes report/presentation copies public while editable sources/audits remain local.","url":"https://raw.githubusercontent.com/PhorkNorak/Explainable-AI-for-Automated-Grading-of-Khmer-Language-Short-Answer-Questions-in-Education/90f127cc160b3e73aacb9b7c2977295332c9aac6/docs/reports/README.md","kind":"project-documentation","is_primary":1}]},{"slug":"korinek-economic-research-agents-guide","title":"AI Agents for Economic Research — Korinek guide","summary":"Anton Korinek’s August 2025 guide introduces AI agents for economic research with worked instructions for literature reviews, econometric code and data analysis.","resource_type":"tutorial-template","disciplines":["business","economics"],"tasks":["data-analysis","literature-research","research-coding","workflow-orchestration"],"evidence_level":"institutional-report","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"The reviewed item is the 2025 update, not proof that the new material itself received the original JEL article’s peer review. Claims about rapid implementation are author claims; examples were not run and model APIs may change.","adoption_requirements":"Access to the guide, suitable model/tool services and an economist able to validate code, data and inference.","updated_at":"2026-10-16T14:58:30Z","url":"https://research.wusong.net/cases/korinek-economic-research-agents-guide","sources":[{"label":"[PDF] AI Agents for Economic Research: August 2025 Update to “Generative ...","url":"https://www.aeaweb.org/content/file?id=23290","kind":"institutional-report","is_primary":1}]},{"slug":"kosmos","title":"Kosmos","summary":"Kosmos coordinates literature and data-analysis agents through an evolving shared world model. The authors present seven discoveries and report expert checks of 102 statements from three reports, with 79.4% supported. These checks provide partial external validation within the study, not independent reproduction of the full system. Human-time equivalents are estimates; results still require domain-expert scrutiny.","resource_type":"workflow-process","disciplines":["multidisciplinary"],"tasks":["data-analysis","hypothesis-generation","literature-research"],"evidence_level":"preprint","outcome_verification":"externally-validated","workflow_level":"L3","workflow_note":"L3 applies only to the researcher-objective literature and processed-data analysis workflow. A scientist supplies a dataset and research objective and evaluates/interprets the resulting report. No intermediate human interaction is claimed. The world model coordinates data-analysis/literature trajectories and revised hypotheses; current data-size/access constraints remain. Twenty-cycle examples and runs up to twelve hours do not establish correct findings. This assessment is limited to documented researcher-directed analysis, not general autonomous science or full independent replication.","source_checked_date":"2026-10-19","limitations":"Study-recruited external statement checks are partial validation, not independent replication of the whole system. Direction sensitivity, roughly 5GB dataset limit and expensive human interpretation remain. Public trajectories/results do not establish full agent-code availability.","adoption_requirements":"Untested suggestion: inspect the pinned sources and validate data, evaluator assumptions, implementation availability and researcher controls before planning a bounded trial. No setup or trial was executed.","updated_at":"2026-10-20T08:01:56Z","url":"https://research.wusong.net/cases/kosmos","sources":[{"label":"Primary manuscript — 2511.02824v2","url":"https://arxiv.org/html/2511.02824v2","kind":"arxiv","is_primary":1}]},{"slug":"labrobfail-laboratory-failure-analysis","title":"LabRobFail: failure-analysis benchmark for chemical laboratory robots","summary":"LabRobFail is a failure-centric simulation, dataset and benchmark framework for robotic failure analysis in chemical laboratories, paired with a specialized vision-language model for structured diagnoses and recovery instructions.","resource_type":"benchmark-dataset","disciplines":["chemistry","robotics"],"tasks":["agent-evaluation","experiment-execution"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Workflow level intentionally unspecified for a reusable component or conceptual reference; researcher oversight and task-specific validation remain necessary.","source_checked_date":"2026-10-17","limitations":"Abstract-only review of arXiv v2. The linked SciRobo repository was identified as an alias but was not inspected. Dataset size, model scores and task-success improvements are author-reported, not audited or reproduced. Simulation and seen-environment results do not establish real chemical safety.","adoption_requirements":"The full benchmark, simulator and model requirements must be inspected separately. Real chemical experiments require expert safety assessment and should not rely on an abstract-level model claim.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/labrobfail-laboratory-failure-analysis","sources":[{"label":"LabRobFail: A Benchmark for Robotic Failure Analysis in Chemical Self-driving Laboratory","url":"https://arxiv.org/abs/2607.23704","kind":"arxiv","is_primary":1}]},{"slug":"langgraph","title":"LangGraph","summary":"LangGraph is an orchestration runtime for stateful agents that can combine deterministic and LLM-driven steps.","resource_type":"framework","disciplines":["multidisciplinary"],"tasks":["research-coding","workflow-orchestration"],"evidence_level":"project-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"General infrastructure rather than a validated research workflow. No benchmark, research outcome or safe autonomous execution is established by this overview.","adoption_requirements":"Python development environment, explicit workflow design, chosen model/tools and permission boundaries.","updated_at":"2026-10-16T14:58:30Z","url":"https://research.wusong.net/cases/langgraph","sources":[{"label":"LangGraph overview - Docs by LangChain","url":"https://docs.langchain.com/oss/python/langgraph/overview","kind":"project-documentation","is_primary":1}]},{"slug":"lattereview","title":"LatteReview: configurable multi-agent literature screening","summary":"Python framework documenting customizable AI reviewers and multi-round literature-review workflows, including title/abstract screening and data abstraction. Reviewers can be arranged in parallel or sequential rounds using different model providers. ","resource_type":"framework","disciplines":["multidisciplinary"],"tasks":["data-extraction","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"L2 applies to the concrete human-configured literature-screening example: people set inclusion/exclusion criteria, models, input fields and round/filter connections; two AI reviewers screen, then a third AI reviewer handles disagreements before result export. The named senior expert is an AI model, not human approval. Final human domain sign-off is not a documented enforced gate; it remains a recommended research safeguard. Fixed configured rounds do not establish autonomous tool planning or a research-iteration loop. Documentation was inspected, not executed or independently evaluated.","source_checked_date":"2026-10-17","limitations":"README-only review, not paper/full documentation or code execution. The README reports benchmark AUC and costs, but their datasets, calculations and generalizability were not adjudicated here and no numeric performance claim is adopted. AI reviewer roles are not human expertise. Researchers remain responsible for screening recall and source fidelity. Model availability and APIs may change.","adoption_requirements":"Python 3.10+, input articles and explicit inclusion/exclusion criteria; API credentials and budget for chosen hosted models, or a compatible local provider. Evaluate provider data policies before sending unpublished or restricted material. Commercial use requires licence review.","updated_at":"2026-10-17T21:22:00Z","url":"https://research.wusong.net/cases/lattereview","sources":[{"label":"PouriaRouzrokh/LatteReview primary documentation","url":"https://github.com/PouriaRouzrokh/LatteReview","kind":"code-repository","is_primary":1}]},{"slug":"lean-copilot","title":"Lean Copilot","summary":"Open-source toolkit allowing language models to be used natively in Lean for proof automation, suggesting tactics/premises and searching for proofs.[11]","resource_type":"tool","disciplines":["computer-science","mathematics"],"tasks":["proof-search"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"A proof assistant checks formal statements, not whether the formalization matches the intended mathematics; no new theorem claim here. Documentation review only - no software execution or independent performance validation.","adoption_requirements":"Compatible Lean project, Git LFS, local or hosted model; build requirements vary by platform.","updated_at":"2026-10-17T06:33:49Z","url":"https://research.wusong.net/cases/lean-copilot","sources":[{"label":"Lean Copilot - repository (lean-dojo/LeanCopilot)","url":"https://github.com/lean-dojo/LeanCopilot","kind":"code-repository","is_primary":1}]},{"slug":"lhonopenclaw-research-task-coordination","title":"LHONOpenClaw research-task coordination","summary":"A documented OpenClaw research skill and task-coordination project for LHON research support. The repository lists funding discovery, researcher mapping, adjacent-field surveys and compilation of research information, with findings submitted through GitHub Issues.","resource_type":"agent-skill","disciplines":["biomedicine"],"tasks":["knowledge-synthesis","literature-research","workflow-orchestration"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation is unspecified; researchers select tasks, review sources, validate outputs and authorize consequential actions.","source_checked_date":"2026-10-17","limitations":"README-only review; skill, task feed and agent execution were not tested. Research acceleration and curing LHON are project aims, not verified outcomes. No medical efficacy or population estimates are endorsed; external submissions and researcher contacts require oversight.","adoption_requirements":"An OpenClaw-compatible skill host and review of the linked skill and tasks before enabling execution. Preserve source citations, respect publication/data rights and avoid patient-identifiable information; obtain approval before public submissions or outreach.","updated_at":"2026-10-17T20:06:54Z","url":"https://research.wusong.net/cases/lhonopenclaw-research-task-coordination","sources":[{"label":"LHONOpenClaw research-task coordination","url":"https://github.com/organicoder42/openclawresearch","kind":"code-repository","is_primary":1}]},{"slug":"lit-review-agent-tools","title":"Awesome AI Literature Review: tool-discovery directory","summary":"Repository-documented tool-discovery directory for AI-assisted literature review, organized into categories spanning search, screening, extraction, synthesis and writing. The README links a searchable interface with workflow-stage, maintenance and licence filters; this is a directory, not an autonomous review engine.","resource_type":"tool","disciplines":["multidisciplinary"],"tasks":["literature-research","workflow-orchestration"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation unspecified. The researcher defines the task, controls access and checks sources and outputs.","source_checked_date":"2026-10-17","limitations":"README-only review; searchable interface, linked tools, health data and recipes were not exercised or independently checked. Inclusion is not an endorsement of accuracy, licence compliance or maintenance. Maintainer claims of executed workflows are not this reviewer's experiments; no comparative strongest-list claim retained.","adoption_requirements":"Browser access suffices for discovery; selected tools have their own costs, credentials, licences and privacy constraints. Verify each upstream project's current terms and evidence before adopting it.","updated_at":"2026-10-17T17:46:50Z","url":"https://research.wusong.net/cases/lit-review-agent-tools","sources":[{"label":"brycewang-stanford/lit-review-agent-tools — primary README","url":"https://github.com/brycewang-stanford/lit-review-agent-tools","kind":"code-repository","is_primary":1}]},{"slug":"literature-derived-materials-data-validation","title":"Validating literature-derived materials data for AI reuse","summary":"A materials-data reliability study traces solid-electrolyte conductivity values from source articles into curated datasets to examine errors relevant to AI training and reuse.","resource_type":"workflow-process","disciplines":["materials-science"],"tasks":["data-extraction","data-management","materials-discovery"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Workflow level left unspecified: the abstract does not establish operational boundaries and researcher responsibilities sufficiently for a level claim. Researchers remain responsible for domain validation, safety and interpretation.","source_checked_date":"2026-10-17","limitations":"Abstract-only primary-source review. No software, hardware, dataset or reported outcome independently reproduced; detailed methods, evaluation settings, distribution and licensing were not inspected. The retrieved record reports July 2026 despite the 2609 identifier prefix; chronology is unresolved and no exact publication date is asserted.","adoption_requirements":"Editorial, untested guidance: retain source-article locations and measurement context alongside extracted values; have a domain researcher resolve unit, axis and text-figure discrepancies before model training. Observe article and dataset reuse rights. No automated validator or dataset download was verified.","updated_at":"2026-10-17T15:13:29Z","url":"https://research.wusong.net/cases/literature-derived-materials-data-validation","sources":[{"label":"When Literature Data Mislead Artificial Intelligence in Materials Discovery","url":"https://arxiv.org/abs/2609.01621","kind":"arxiv","is_primary":1}]},{"slug":"literature-meets-data","title":"Literature Meets Data (hypothesis-generation framework)","summary":"Hypothesis-generation framework combining literature-based insights with data in an LLM-powered pipeline; reported gains over few-shot, literature-only and data-only baselines, with human-evaluation gains on two decision tasks.[14]","resource_type":"framework","disciplines":["computer-science","social-science"],"tasks":["hypothesis-generation","knowledge-synthesis"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans set objectives and validate results.","source_checked_date":"2026-10-16","limitations":"Author-reported results; not independently reproduced (including the human evaluation). Published version checked at record/abstract level.","adoption_requirements":"LLM access; literature corpus and task datasets; no packaged implementation verified.","updated_at":"2026-10-17T08:00:32Z","url":"https://research.wusong.net/cases/literature-meets-data","sources":[{"label":"Literature Meets Data: A Synergistic Approach to Hypothesis Generation - arXiv:2410.17309","url":"https://arxiv.org/abs/2410.17309","kind":"arxiv","is_primary":1}]},{"slug":"literature-review-zh","title":"literature-review-zh: humanities and social-science review skill","summary":"AI-assistant skill documented for literature-review writing with humanities and social-science adaptations: discipline routing, citation tracing, historical-source handling and type-specific reference verification. Its README describes version and translator attribution, secondary-citation marking and checks against independent factual sources.","resource_type":"agent-skill","disciplines":["humanities","social-science"],"tasks":["literature-research","scientific-writing"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation unspecified. The researcher defines the task, controls access and checks sources and outputs.","source_checked_date":"2026-10-17","limitations":"README-only review; no execution or citation-accuracy evaluation. The maintainer says the original skill author and licence are unknown; MIT covers only their modifications/additions, not unidentified upstream content. The README reports two mentioned helper scripts absent (repository tree/code not independently inspected), with a documented fallback to search tools or user-supplied results. Missing upstream identity prevents exhaustive derivation/alias resolution; no relationship to queue 201 is asserted. Admission is a bounded documentary listing, not a runnable-package verification or rights clearance.","adoption_requirements":"Requires a SKILL.md-compatible assistant and bundled locales/references, access to relevant catalogues or user-provided search results, and researcher checks of editions, page numbers and attribution. Resolve rights for the intended reuse; do not presume the entire inherited skill is MIT-licensed. Keep restricted archives and private documents within approved systems.","updated_at":"2026-10-17T17:46:50Z","url":"https://research.wusong.net/cases/literature-review-zh","sources":[{"label":"Darry-cpu/literature-review-zh — primary README","url":"https://github.com/Darry-cpu/literature-review-zh","kind":"code-repository","is_primary":1}]},{"slug":"llm-assisted-microscopy-api-workflows","title":"LLM-assisted microscopy: translating expert ideas into API workflows","summary":"An author-reported study of ChatGPT4 as a language interface for scanning-probe microscopy, combining in-house and commercial APIs for experimental design, workflow programming and generic microscopy-image analysis.","resource_type":"workflow-process","disciplines":["computer-science","materials-science","physics"],"tasks":["experiment-design","image-analysis","research-coding"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Resource-level workflow level intentionally unspecified. Conceptual references and reusable process patterns do not themselves establish an autonomous workflow; researchers retain task-specific interpretation and oversight.","source_checked_date":"2026-10-17","limitations":"ArXiv abstract-only review for the workflow description; official ORNL journal metadata additionally reviewed, not the journal methods or supplements. The official record identifies a peer-reviewed Machine Learning: Science and Technology article (2024), DOI 10.1088/2632-2153/ad52e9. Direct publisher retrieval hit a bot challenge; a web-extracted publisher excerpt was read but is not cited as complete methods evidence. No code, microscope, prompts, accuracy, runtime, safety outcome or fine-tuning result was reproduced. Generic image analysis must not be upgraded to expert diagnostic interpretation.","adoption_requirements":"The described workflow uses microscope-control APIs and human scientific expertise. Editorial, untested adoption also requires authorised instrument access, API documentation, code inspection, privacy safeguards and laboratory-specific safety controls; exact dependencies and costs were not established from the abstract.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/llm-assisted-microscopy-api-workflows","sources":[{"label":"Human Expertise and AI Efficiency in Microscopy — arXiv v1 abstract","url":"https://arxiv.org/abs/2401.13803","kind":"arxiv","is_primary":1},{"label":"ORNL official publication record — journal and peer-review metadata","url":"https://impact.ornl.gov/en/publications/synergizing-human-expertise-and-ai-efficiency-with-language-model","kind":"institutional-report","is_primary":1}]},{"slug":"llm-authorship-senior-author-analogy","title":"LLM-assisted authorship: the senior-author analogy","summary":"A conceptual authorship-ethics reference arguing that some human uses of LLM-generated research drafts can be analogous to senior authorship, where project guidance and responsibility may count without direct writing.","resource_type":"tutorial-template","disciplines":["humanities","medicine","multidisciplinary"],"tasks":["scientific-writing"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation classification unspecified. This is a research/reference listing, not an exercised workflow; researchers select objectives and check interpretation and outputs.","source_checked_date":"2026-10-17","limitations":"Abstract-only substantive review, corroborated by Oxford’s official preprint record, which explicitly says not peer reviewed. The argument is conditional and contested, not an empirical outcome, institutional policy or a blanket permission to publish LLM-written work. Detailed conditions and counterarguments were not inspected; this is an ethics-discussion reference, not an operational authorship checklist or legal advice. No writing system, authorship decision or effectiveness experiment was executed.","adoption_requirements":"Editorial, untested use: compare the argument with the relevant journal, institution and disciplinary authorship/disclosure rules. Read the full conditions before applying the analogy to a real contributor; assess accountability and confidentiality with the research team.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/llm-authorship-senior-author-analogy","sources":[{"label":"Authorship Without Writing — primary arXiv abstract","url":"https://arxiv.org/abs/2509.05390","kind":"arxiv","is_primary":1},{"label":"Oxford ORA — official preprint and peer-review-status record","url":"https://ora.ox.ac.uk/objects/uuid:7b3c28a7-aa61-4ca9-a665-758aed5cbccd","kind":"institutional-report","is_primary":1}]},{"slug":"llm-in-the-loop","title":"LLM-in-the-loop: Leveraging Large Language Model for Thematic Analysis","summary":"Peer-reviewed framework for thematic analysis in which a human and an LLM collaborate via in-context learning to generate a codebook.[18]","resource_type":"framework","disciplines":["multidisciplinary","social-science"],"tasks":["data-analysis","research-coding"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-16","limitations":"Two case studies only (survey datasets); coding quality reported similar to human coders but not independently reproduced. Abstract-level review only - full-text methods and any released materials were not inspected, and no independent performance validation.","adoption_requirements":"LLM access for in-context prompting (paper uses GPT-3.5 as an example) and qualitative research expertise to assess generated codebooks.","updated_at":"2026-10-17T06:33:49Z","url":"https://research.wusong.net/cases/llm-in-the-loop","sources":[{"label":"LLM-in-the-loop: Leveraging Large Language Model for Thematic Analysis","url":"https://arxiv.org/abs/2310.15100","kind":"arxiv","is_primary":1}]},{"slug":"llm-laboratory-instrumentation-control","title":"LLM-assisted laboratory instrumentation control","summary":"Case study of ChatGPT-assisted custom control scripting for a laboratory setup usable as a single-pixel camera or scanning photocurrent microscope.","resource_type":"use-case","disciplines":["engineering","physics"],"tasks":["experiment-execution","research-coding"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":"L2","workflow_note":"L2 conservatively describes the bounded AI-assisted application in the abstract; no end-to-end autonomous scientific discovery or safety guarantee is asserted. Researchers retain responsibility for validation and safe use.","source_checked_date":"2026-10-17","limitations":"Abstract-only review of arXiv v1. The record identifies an accepted manuscript in Small Structures (2025) and links supporting data/code, but the journal body and artifact contents were not inspected. No quantitative time saving, code reliability, unattended safety or generalization is established here. No software or equipment was run. The retrieved record states 25 March 2026 despite the 2604 identifier prefix; chronology remains unresolved. The displayed 2025 year follows its journal citation, not an inferred arXiv date.","adoption_requirements":"Untested editorial guidance: requires instrumentation expertise, appropriate equipment and interfaces, model access, code inspection and supervised testing with physical safety interlocks. Check data rights and provider policies before sharing laboratory details.","updated_at":"2026-10-17T15:13:29Z","url":"https://research.wusong.net/cases/llm-laboratory-instrumentation-control","sources":[{"label":"Toward Full Autonomous Laboratory Instrumentation Control with Large Language Models — primary arXiv record","url":"https://arxiv.org/abs/2604.03286","kind":"arxiv","is_primary":1}]},{"slug":"llm-thematic-analysis","title":"LLM-assisted inductive thematic analysis of research interviews","summary":"An exploratory study uses GPT-3.5-Turbo to partially reproduce inductive thematic-analysis phases on two existing open-access interview datasets, comparing outputs with previous human analyses.","resource_type":"use-case","disciplines":["social-science"],"tasks":["data-analysis"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":"L2","workflow_note":"Human-orchestrated qualitative analysis across coding and theme development, partially following the six thematic-analysis phases. The researcher supplies data, prompts and interpretation and compares outputs with earlier analyses; no autonomous tool-using agent is demonstrated.","source_checked_date":"2026-10-15","limitations":"The study is explicitly exploratory and does not establish formal procedures. LLM pattern generation is not human interpretation; generalization, model variability and reflexivity require care.","adoption_requirements":"Appropriate consent and data-governance review before using an LLM, qualitative-method expertise, recorded prompts and model settings, and human checking against the original interviews.","updated_at":"2026-10-15T17:25:09Z","url":"https://research.wusong.net/cases/llm-thematic-analysis","sources":[{"label":"De Paoli, Performing an Inductive Thematic Analysis of Semi-Structured Interviews With a Large Language Model: An Exploration and Provocation on the Limits of the Approach, Social Science Computer Review (2024)","url":"https://doi.org/10.1177/08944393231220483","kind":"doi","is_primary":1}]},{"slug":"llm-thematic-analysis-empirical-legal-studies","title":"LLM-assisted thematic analysis in empirical legal studies","summary":"Study of legal-expert collaboration with GPT-4 for initial coding, theme discovery and classification of theft-related court facts, evaluated by the authors on 785 descriptions.","resource_type":"use-case","disciplines":["law","social-science"],"tasks":["data-analysis"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":"L2","workflow_note":"Defined researcher-directed coding/theme-analysis workflow; human interpretation and checking remain necessary. No autonomous research loop established.","source_checked_date":"2026-10-17","limitations":"Results concern one theft-related court-opinion dataset and GPT-4; transfer to other jurisdictions, offences or models is not established. Primary-source abstract and metadata review only; full-text methods, implementation and data were not inspected. No model or experiment was run and no outcome independently reproduced.","adoption_requirements":"For adaptation, access to a suitable LLM, lawfully usable court-text data and a legal researcher able to check codes, themes and contextual meaning; prompts and implementation were not reproduced.","updated_at":"2026-10-17T11:14:58Z","url":"https://research.wusong.net/cases/llm-thematic-analysis-empirical-legal-studies","sources":[{"label":"Using Large Language Models to Support Thematic Analysis in Empirical Legal Studies — arXiv","url":"https://arxiv.org/abs/2310.18729","kind":"arxiv","is_primary":1}]},{"slug":"llm-thematic-analysis-italian-interviews","title":"LLM thematic analysis of Italian interviews","summary":"Targeted study of LLM thematic analysis using Italian-language interview data and Italian prompts, comparing themes with prior researcher analyses.","resource_type":"use-case","disciplines":["linguistics","social-science"],"tasks":["data-analysis"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":"L2","workflow_note":"Defined researcher-directed coding/theme-analysis workflow; human interpretation and checking remain necessary. No autonomous research loop established.","source_checked_date":"2026-10-17","limitations":"A targeted Italian-language test does not establish effectiveness across other languages or cultures; thematic resemblance is an author interpretation, not independently measured here. Primary-source abstract and metadata review only; full-text methods, implementation and data were not inspected. No model or experiment was run and no outcome independently reproduced.","adoption_requirements":"For adaptation, an Italian-capable LLM, authorized interview data and Italian-speaking qualitative researchers to evaluate meaning and context. Exact model settings and setup were not checked.","updated_at":"2026-10-17T11:14:58Z","url":"https://research.wusong.net/cases/llm-thematic-analysis-italian-interviews","sources":[{"label":"Thematic Analysis with Large Language Models: does it work with languages other than English? A targeted test in Italian — arXiv","url":"https://arxiv.org/abs/2404.08488","kind":"arxiv","is_primary":1}]},{"slug":"llm4sr-research-cycle-survey","title":"LLM4SR: survey of LLM roles across the scientific research cycle","summary":"The LLM4SR preprint surveys LLM-supported methods and evaluation benchmarks across four broad stages of scientific research.","resource_type":"framework","disciplines":["computer-science","multidisciplinary"],"tasks":["literature-research","workflow-orchestration"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level assigned. The abstract maps LLM methods and evaluation topics across hypothesis discovery, experiment planning/implementation, writing and peer review. It does not establish a single connected workflow, researcher control, output validation or operational loop.","source_checked_date":"2026-10-19","limitations":"Official abstract only. The full survey, search/inclusion procedure, benchmarks, linked repository and cited studies were not audited. Claims of being first, systematic, comprehensive or revolutionary remain author descriptions. No model, vendor capability, scientific result or outcome across surveyed systems was verified. No paper-to-repository linkage is inferred.","adoption_requirements":"Untested suggestion: choose one stage and inspect the relevant underlying primary studies, coverage and evaluation limits before reuse.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/llm4sr-research-cycle-survey","sources":[{"label":"Primary survey abstract","url":"https://arxiv.org/abs/2501.04306","kind":"arxiv","is_primary":1}]},{"slug":"longsizhuo-openinvest","title":"openInvest","summary":"Repository-documented investment research framework combining separated analyst roles, cross-challenge and recorded decisions with a backtesting harness. Listed for studying decision processes, not as investment advice or a profitable strategy.","resource_type":"framework","disciplines":["business","economics"],"tasks":["agent-evaluation","data-analysis"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation unspecified. Researchers approve objectives and inspect outputs; documentation is not an executed workflow.","source_checked_date":"2026-10-17","limitations":"README-only source review by an AI editorial agent; no implementation execution, independent outcome evaluation, security audit or full linked-document review. HEAD is mutable; saved content and hashes preserve the inspected version. README contains negative author-reported tests and explicitly disclaims alpha; no quantitative outcome is included in the listing. Its cutoff explanation is not endorsed: historical backtesting can still suffer model-training leakage and requires independent analysis. Public state commits may disclose private finances; do not follow automatic deployment instructions without privacy review.","adoption_requirements":"Python environment, agent or MCP/CLI client, market data and protected portfolio state; automated modes may require provider credentials. Verify data rights, model costs and permissions before adoption.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/longsizhuo-openinvest","sources":[{"label":"longsizhuo/openInvest — README","url":"https://raw.githubusercontent.com/longsizhuo/openInvest/HEAD/README.md","kind":"code-repository","is_primary":1}]},{"slug":"loyalty-fraud-ml-conceptual-review","title":"Loyalty fraud detection: conceptual ML comparison framework","summary":"Conceptual review of machine-learning approaches to loyalty-program fraud. The abstract frames loyalty abuse as a distinct analytical problem and discusses hybrid pipelines, graph-based modeling and behavior-aware scoring as methodological directions, rather than reporting a deployed system.","resource_type":"framework","disciplines":["business","computer-science"],"tasks":["data-analysis","knowledge-synthesis"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation is unspecified. Researchers choose questions, evaluation design and decisions; documentation review is not workflow execution.","source_checked_date":"2026-10-17","limitations":"Abstract-only publisher review; no benchmark data, operational deployment, comparative accuracy or independent outcome reproduction verified. Recommendations are conceptual/author interpretations. Publisher labels the journal peer reviewed; editorial quality was not independently audited.","adoption_requirements":"Editorial, untested guidance: read the underlying paper and assess local fraud labels, access rights, privacy and evaluation design before selecting a modeling approach. No deployable code or validated workflow is established by this review.","updated_at":"2026-10-17T18:33:50Z","url":"https://research.wusong.net/cases/loyalty-fraud-ml-conceptual-review","sources":[{"label":"A Systematic Review of Machine Learning Approaches For AI-Driven Fraud Detection in Loyalty Programs","url":"https://aimjournals.com/index.php/ijidml/article/view/612","kind":"doi","is_primary":1}]},{"slug":"machine-learning-two-dimensional-topological-insulators","title":"Machine learning for materials discovery: two-dimensional topological insulators","summary":"A machine-learning materials-screening strategy uses databases of ab initio calculations to predict electronic topology and guide searches for two-dimensional topological insulators.","resource_type":"workflow-process","disciplines":["materials-science","physics"],"tasks":["data-analysis","materials-discovery"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Automation level unspecified: a resource-level description is not a measured autonomous workflow. Researchers select goals and data and validate outputs.","source_checked_date":"2026-10-17","limitations":"Primary arXiv abstract-only review; full text and software not reviewed or executed. The candidate counts and DFT corroboration are author-reported computational results, not physical synthesis or editorial reproduction. Dataset composition, splits and robustness were not checked. The arXiv record lists a journal reference; the publisher article was not reviewed.","adoption_requirements":"Editorial, untested guidance: inspect the paper and supplementary materials before selecting databases, model inputs and DFT settings. Obtain appropriate data permissions and physics expertise; no runnable recipe was tested.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/machine-learning-two-dimensional-topological-insulators","sources":[{"label":"Machine learning for materials discovery: two-dimensional topological insulators — arXiv 2107.07028","url":"https://arxiv.org/abs/2107.07028","kind":"arxiv","is_primary":1}]},{"slug":"madore-archival-ai-research-slides","title":"Beyond Keywords: June15 archival research method slides (Madore)","summary":"The selected June15 Beyond Keywords presentation documents user-scoped archival questions, model-selected read-only tools and attributed historical synthesis; the original slides-collection identity is retained.","resource_type":"tutorial-template","disciplines":["humanities","social-science"],"tasks":["archival-research","knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L3","workflow_note":"Scoped slide-described L3 for the selected June15 Beyond Keywords archival research method: user question/depth, model-selected read-only tools and scoping/searching/reading/triangulation/attributed synthesis. It is not a level for the entire presentation collection, a runtime audit, enforced expert signoff or L4 scientific experimentation.","source_checked_date":"2026-10-19","limitations":"Documentation only; no repository code, model calls, installation, data processing or experiment executed. Primary API/raw source retrieval does not validate implementation, security, performance or scientific outcomes. One selected deck plus manifest and selectedREADME context; no audit of allseven decks or embedded190line skill. Saved-run screenshots/chats were not inspected and illustrative search slices are not validated retrieval experiments. The deck’s blanket no-AI-at-query-time statement is qualified by its explicit live Gemini query-embedding tool. No inherited IWAC outcomes, collection counts, universal multilingual accuracy or enforced expert signoff claimed. The workshop/collective-position-paper material elsewhere in the collection is outside this selected June15 method scope.","adoption_requirements":"Read the public slides; any later underlying tool trial needs its own current setup/provider/data-rights review. The presentation is not the installed MCP server or full embeddedskill.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/madore-archival-ai-research-slides","sources":[{"label":"fmadore/slides beyond-keywords.json","url":"https://raw.githubusercontent.com/fmadore/slides/675857512df75db1b5810e52e2b648c2a2a66614/talks/2026-06-15-luxembourg-beyond-keywords/index.html","kind":"project-documentation","is_primary":1},{"label":"fmadore/slides talk-manifest-response.json","url":"https://raw.githubusercontent.com/fmadore/slides/675857512df75db1b5810e52e2b648c2a2a66614/talks/talks.json","kind":"project-documentation","is_primary":1}]},{"slug":"malcom-ai-research-assistant","title":"Malcom: document-context and web-research workspace","summary":"Malcom documents an AI workspace that combines uploaded-document context with web search, retrieved source excerpts and retained source links for model-assisted research.","resource_type":"tool","disciplines":["multidisciplinary"],"tasks":["knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation is unspecified. Researchers define the task, inspect sources/code and validate outputs; this review did not execute a workflow.","source_checked_date":"2026-10-17","limitations":"README-only source review; application, search quality, document handling, privacy controls and tests were not exercised. Source links do not guarantee correct citations. No independent outcome evaluation or license audit.","adoption_requirements":"README prerequisites: Node.js 20 or newer, npm and an Ollama-compatible chat server. Optional Supabase, Stripe, Brave Search and Redis are not required for the documented local-development route. Review storage and provider privacy before using confidential documents.","updated_at":"2026-10-17T17:46:50Z","url":"https://research.wusong.net/cases/malcom-ai-research-assistant","sources":[{"label":"Malcom: document-context and web-research workspace — primary documentation","url":"https://gitlab.com/0xmuditAsia/malcom-ai-research-assistant/-/raw/main/README.md","kind":"code-repository","is_primary":1}]},{"slug":"mammography-roi-classification-review","title":"Mammography ML review: ROI extraction and classification comparison","summary":"Abstract-level methodological review of mammographic image classification. The authors describe comparison of ROI extraction strategies, machine-learning methods and datasets, with accuracy and frequency-based statistics, across a stated selection of 60 papers.","resource_type":"workflow-process","disciplines":["computer-science","medicine"],"tasks":["image-analysis","knowledge-synthesis"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation is unspecified. Researchers choose questions, evaluation design and decisions; documentation review is not workflow execution.","source_checked_date":"2026-10-17","limitations":"Abstract-only publisher review; selection criteria, datasets, per-model results and comparative superiority are not verified. The count of 60 papers is author-reported. Journal presents itself as peer-reviewed; review quality was not independently audited. This is not a clinically validated diagnostic tool.","adoption_requirements":"Editorial, untested guidance: consult the paper before designing an ROI classification comparison; independently define data splits, inclusion criteria and evaluation metrics. Clinical data require governance, consent/access authorization and specialist oversight.","updated_at":"2026-10-17T18:33:50Z","url":"https://research.wusong.net/cases/mammography-roi-classification-review","sources":[{"label":"Systematic Review on Machine Learning and Deep Learning Approaches for Mammography Image Classification","url":"https://doi.org/10.5373/jardcs/v12i7/20202015","kind":"doi","is_primary":1}]},{"slug":"master-materials-simulation-reasoning","title":"MASTER: multi-agent reasoning for materials simulations","summary":"MASTER is an active-learning framework in which large language models design, execute and interpret atomistic simulations for materials research, according to the authors.","resource_type":"framework","disciplines":["materials-science"],"tasks":["experiment-design","materials-discovery","workflow-orchestration"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Workflow level left unspecified: the abstract does not establish operational boundaries and researcher responsibilities sufficiently for a level claim. Researchers remain responsible for domain validation, safety and interpretation.","source_checked_date":"2026-10-17","limitations":"Abstract-only primary-source review. No software, hardware, dataset or reported outcome independently reproduced; detailed methods, evaluation settings, distribution and licensing were not inspected.","adoption_requirements":"Editorial, untested guidance: assess access to density-functional-theory software and compute, model access and electronic-structure expertise. Check simulation setup and convergence, inspect proposed chemistry and account for data privacy and compute cost. A runnable distribution was not verified.","updated_at":"2026-10-17T15:13:29Z","url":"https://research.wusong.net/cases/master-materials-simulation-reasoning","sources":[{"label":"Hierarchical Multi-agent Large Language Model Reasoning for Autonomous Functional Materials Discovery","url":"https://arxiv.org/abs/2512.13930","kind":"arxiv","is_primary":1}]},{"slug":"materials-generative-model-representation-review","title":"Materials generative models: representation and integration reference","summary":"Selected review sections compare material representations and discuss integrating generative predictions with experiments.","resource_type":"framework","disciplines":["chemistry","materials-science"],"tasks":["experiment-design","materials-discovery"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level assigned. The reference compares materials representations and experimental integration, but does not establish a single researcher-controlled workflow or output-acceptance protocol. Model names and closed-loop language do not determine autonomy.","source_checked_date":"2026-10-19","limitations":"Selected retained HTML sections on representation, integration and limitations only; not the full paper, references, algorithms, data or outcomes. Some extracted reference tokens are malformed. Architecture, application and numerical-gain claims were not promoted as verified facts. No universal representation, physical validity, superiority or validated closed-loop discovery is established; recommendations are not demonstrated enforcement.","adoption_requirements":"Untested suggestion: check representation suitability, data bias, physical validity, computational costs and the chosen underlying primary methods before reuse.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/materials-generative-model-representation-review","sources":[{"label":"Selected representation, integration and limitation sections","url":"https://arxiv.org/html/2508.03278","kind":"arxiv","is_primary":1}]},{"slug":"mattergen","title":"MatterGen","summary":"Generative model for inorganic material design with property-constrained fine-tuning.[5]","resource_type":"tool","disciplines":["chemistry","materials-science"],"tasks":["materials-discovery"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Generated structures still require physical validation; the repository states restricted ICSD data cannot be shared and replication may differ. No software execution or independent performance validation.","adoption_requirements":"Model checkpoints and Python environment; the documented standard setup assumes Linux/CUDA; Git LFS for data.","updated_at":"2026-10-17T06:33:49Z","url":"https://research.wusong.net/cases/mattergen","sources":[{"label":"MatterGen - repository (microsoft/mattergen)","url":"https://github.com/microsoft/mattergen","kind":"code-repository","is_primary":1}]},{"slug":"maxqda-ai-assist","title":"MAXQDA AI Assist","summary":"MAXQDA AI Assist offers qualitative-data summaries, paraphrases, code suggestions and document conversations.","resource_type":"tool","disciplines":["business","humanities","social-science"],"tasks":["data-analysis","knowledge-synthesis"],"evidence_level":"vendor-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Vendor feature documentation, not independent evidence of coding quality or saved time. Interpretive judgments, confidentiality and code consistency need human review.","adoption_requirements":"MAXQDA with applicable AI Assist access, approved qualitative data and a researcher-owned coding protocol.","updated_at":"2026-10-16T14:58:30Z","url":"https://research.wusong.net/cases/maxqda-ai-assist","sources":[{"label":"What is MAXQDA AI Assist?","url":"https://help.maxqda.com/en/support/solutions/articles/80001135283-what-is-maxqda-ai-assist-","kind":"vendor-documentation","is_primary":1}]},{"slug":"mechagents","title":"MechAgents","summary":"MechAgents uses collaborating language-model roles to formulate, execute and correct finite-element analyses. The paper illustrates elasticity and hyperelasticity tasks, with larger teams adding planning and criticism. These are bounded author demonstrations rather than a general success-rate benchmark. Execution alone cannot establish a physically correct formulation. Public code is available and was not run in this review.","resource_type":"framework","disciplines":["engineering"],"tasks":["data-analysis","research-coding"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":"L3","workflow_note":"L3 applies only to the documented researcher-assigned finite-element demonstrations. Humans assign mechanics tasks and the originating authors inspect formulations/results; user-proxy/executor roles are software roles, not proof of a human approval at every turn. Planning, code execution and mutual criticism can repair bounded elasticity demonstrations; silent formulation errors remain possible. Conversation/context and assigned problems bound the examples; no general mechanics success rate, safety protocol or physical experimental validation is claimed.","source_checked_date":"2026-10-19","limitations":"Bounded FEniCS demonstrations; silent physical/formulation errors can persist despite execution. General mechanics accuracy and physical experimental validation are not shown.","adoption_requirements":"Untested suggestion: inspect the pinned sources and validate data, evaluator assumptions, implementation availability and researcher controls before planning a bounded trial. No setup or trial was executed.","updated_at":"2026-10-20T08:01:56Z","url":"https://research.wusong.net/cases/mechagents","sources":[{"label":"Primary manuscript — 2311.08166v1","url":"https://arxiv.org/pdf/2311.08166v1","kind":"arxiv","is_primary":1},{"label":"Primary project repository or artifact (version limits apply)","url":"https://github.com/lamm-mit/MechAgents/blob/4d02b14e16fc1edcdce32e11d3c0a66abdfa2fbc/README.md","kind":"code-repository","is_primary":1}]},{"slug":"megadetector","title":"MegaDetector","summary":"Camera-trap image detection model that finds animals, people and vehicles, deliberately stopping short of species identification.[8]","resource_type":"tool","disciplines":["biology"],"tasks":["image-analysis"],"evidence_level":"project-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Detection is not species identification; false detections and missed images require downstream review. Documentation review only - no software execution or independent performance validation.","adoption_requirements":"Camera-trap images; Python environment; downstream species classification and validation.","updated_at":"2026-10-17T06:33:49Z","url":"https://research.wusong.net/cases/megadetector","sources":[{"label":"MegaDetector - official documentation (Microsoft)","url":"https://microsoft.github.io/MegaDetector/","kind":"project-documentation","is_primary":1}]},{"slug":"meghaka-trustworthy-agentic-ai-healthcare","title":"Trustworthy Agentic AI for Healthcare — research prototype","summary":"A repository-documented NHANES-based research prototype combining diabetes-related early-risk modelling, explanations, trust and fairness evaluation, safety gates and a multi-agent workflow. The README describes a locked logistic-regression model and a Streamlit demonstrator.","resource_type":"framework","disciplines":["biomedicine"],"tasks":["agent-evaluation","data-analysis"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation is unspecified; researchers select tasks, review sources, validate outputs and authorize consequential actions.","source_checked_date":"2026-10-17","limitations":"README-only evidence; notebooks, model artifacts, clinical validity, fairness and deployed application were not independently evaluated. The project explicitly disclaims diagnosis and treatment recommendations. SHAP exploration and coefficient-based demonstrator explanations are different methods and must not be conflated. This is not a clinically validated system; individual risk calibration, safety and fairness are not established.","adoption_requirements":"Inspect the NHANES source/processing notes, model artifacts and demonstrator dependencies before reuse; repository documentation alone does not establish a safe clinical setup. Respect public-use data conditions and avoid uploading private health records; qualified clinical oversight is required for any interpretation.","updated_at":"2026-10-17T20:06:54Z","url":"https://research.wusong.net/cases/meghaka-trustworthy-agentic-ai-healthcare","sources":[{"label":"Trustworthy Agentic AI for Healthcare — research prototype","url":"https://github.com/MeghaKA/Trustworthy-Agentic-AI-for-Healthcare","kind":"code-repository","is_primary":1}]},{"slug":"mephisto","title":"Mephisto","summary":"Mephisto uses tree-guided galaxy fitting and memory-based refinement. Its author evaluation includes 256 COSMOS2020 galaxies and 31 Little Red Dots; a knowledge comparison uses 208 training and 48 held-out galaxies. Comparable statistical fits do not uniquely establish physical explanations. The cited code archive returned 429 and was not retried; implementation version and independent replication remain unverified.","resource_type":"workflow-process","disciplines":["physics"],"tasks":["data-analysis","optimization"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":"L3","workflow_note":"L3 applies only to the researcher-directed galaxy-fitting search. Researchers supply observations/problem settings; the described language interface permits understanding, validating and guiding decisions. This is not evidence of an expert user study. Tree search proposes CIGALE configurations, runs fits and uses discrepancies/memory for refinement. Configured tree search and data/model bounds limit the method. Similar statistical fits do not resolve physical degeneracy; no unconditional runtime or correctness guarantee.","source_checked_date":"2026-10-19","limitations":"Betterχ²/model agreement is not unique physical truth:AGN/starburst alternatives fit similarly. Grid-size ratios are not universal end-to-end speedups. Historic model/token costs, visual limitations and prepared-data dependence constrain transfer; no independent replication. Journal-formatted text says submitted, not verified acceptance.","adoption_requirements":"Untested suggestion: inspect the pinned sources and validate data, evaluator assumptions, implementation availability and researcher controls before planning a bounded trial. No setup or trial was executed.","updated_at":"2026-10-20T08:01:56Z","url":"https://research.wusong.net/cases/mephisto","sources":[{"label":"Primary manuscript — arXiv2510.08354v1;9October2025;older conference abstract2409.14807 is related, not substituted","url":"https://arxiv.org/html/2510.08354v1","kind":"arxiv","is_primary":1}]},{"slug":"microsoft-agent-framework","title":"Microsoft Agent Framework","summary":"Open, multi-language framework for building production-grade AI agents and multi-agent workflows in .NET and Python; identified by AutoGen maintainers as its successor.[17]","resource_type":"framework","disciplines":["computer-science","multidisciplinary"],"tasks":["workflow-orchestration"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Generic infrastructure only; applicability is inferred from orchestration functionality, not proven research gains. Distinct successor project, not a second AutoGen use-case. Documentation review only - no software execution or independent performance validation.","adoption_requirements":"Python/.NET integration, model access, research-specific tools, permissions and evaluation.","updated_at":"2026-10-17T06:33:49Z","url":"https://research.wusong.net/cases/microsoft-agent-framework","sources":[{"label":"Microsoft Agent Framework - repository (microsoft/agent-framework)","url":"https://github.com/microsoft/agent-framework","kind":"code-repository","is_primary":1}]},{"slug":"ml-iterator-interview-research-documentation","title":"ML_Iterator_Dare2Dream: documented interview summaries and research-data Q&A","summary":"Documentation describes AI-assisted interview summaries and a Q&A interface for research data.","resource_type":"tool","disciplines":["multidisciplinary","social-science"],"tasks":["data-extraction","knowledge-synthesis"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level assigned. The README describes human-conducted interviews, AI listening/summaries and later research-data Q&A. Researcher output checks, approvals and model/data flow are not sufficiently documented. Only the tool’s stated intent is established.","source_checked_date":"2026-10-19","limitations":"README only. Code, installer/archive, application, manual, recordings, data and models were not audited or executed. Release, documentation and installer links point to the same ZIP; a canonical version, OS-specific distribution and separate manual were not verified. This ambiguity does not prove malware or absence of an application. Local-storage, security, cost and ease claims are unverified; Internet access is required and provider handling is unclear. No participant data was accessed and no consent, privacy or effectiveness assurance is established. Moving HEAD is unpinned and non-atomic.","adoption_requirements":"Untested reference only: clarify distribution provenance, model/data handling, participant consent and researcher output checks before considering a trial; no ZIP installation is recommended here.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/ml-iterator-interview-research-documentation","sources":[{"label":"Project README snapshot","url":"https://raw.githubusercontent.com/blight07262021/ML_Iterator_Dare2Dream/HEAD/README.md","kind":"code-repository","is_primary":1}]},{"slug":"ml-meta-analysis-methodology-reference","title":"ML meta-analysis: baseline-comparison methodology reference","summary":"A methodological position paper proposes systematic synthesis of empirical ML comparisons.","resource_type":"framework","disciplines":["computer-science","multidisciplinary"],"tasks":["experiment-design","knowledge-synthesis"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level assigned. The proposed reference framework aggregates empirical ML baseline comparisons across implementations, settings and datasets. It does not establish AI performing the review or one supervised AI research workflow.","source_checked_date":"2026-10-19","limitations":"Selected PDF passages from a 2018 workshop/preprint, obtained through potentially cached extraction. Raw PDF/version parity, references, data, pooled estimates and effectiveness were not audited. Proposed benefits are not demonstrated outcomes or current-state claims. Statistical pooling does not remove publication bias or make incompatible metrics comparable.","adoption_requirements":"Untested suggestion: check baseline comparability, selection bias, metric definitions and pooling assumptions before designing a synthesis.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/ml-meta-analysis-methodology-reference","sources":[{"label":"Selected methodological paper passages","url":"https://arxiv.org/pdf/1812.01074","kind":"arxiv","is_primary":1}]},{"slug":"modern-computational-methods-drug-repurposing","title":"Modern Computational Methods for Drug Repurposing","summary":"A review overview of computational drug repurposing that distinguishes physicochemical simulations, similarity-based methods using chemical and biological information, and learning-based approaches for therapeutic hypothesis generation.","resource_type":"workflow-process","disciplines":["biomedicine"],"tasks":["drug-repurposing","knowledge-synthesis"],"evidence_level":"project-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation is unspecified; researchers select tasks, review sources, validate outputs and authorize consequential actions.","source_checked_date":"2026-10-17","limitations":"Publisher abstract-only review; not a tested workflow or independently verified therapeutic result. Peer-review process not established from the inspected abstract; no effectiveness, safety, clinical translation or comparative superiority is asserted.","adoption_requirements":"Conceptual orientation requires pharmacology and computational-method knowledge; implementation needs method-specific datasets and validation beyond this abstract. Data rights and clinical oversight remain necessary.","updated_at":"2026-10-17T20:06:54Z","url":"https://research.wusong.net/cases/modern-computational-methods-drug-repurposing","sources":[{"label":"Modern Computational Methods for Drug Repurposing","url":"https://drugrepocentral.scienceopen.com/hosted-document?doi=10.58647/DRUGREPO.26.1.0001","kind":"project-documentation","is_primary":1}]},{"slug":"molt-agentic-rl-framework","title":"Molt: agentic reinforcement-learning framework","summary":"Molt documents an agentic reinforcement-learning stack using Ray for placement and asynchronous queues, vLLM for rollouts, and NVIDIA AutoModel/FSDP2 for training. Researchers define rewards and interactions through Python Env or ChatAgent interfaces.","resource_type":"framework","disciplines":["computer-science"],"tasks":["experiment-execution","optimization","research-coding"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation is unspecified. Researchers choose goals, inputs, permissions and evaluation criteria and remain responsible for interpreting outputs.","source_checked_date":"2026-10-17","limitations":"Partial README review, focused on identity, architecture, installation and source-link boundaries; no code execution, training, hardware validation, paper review or benchmark reproduction. Frontier-scale, speed and convergence claims are not verified or retained. Documentation and branch contents are mutable; the retrieved body is hash-preserved but no commit was resolved.","adoption_requirements":"README specifies a CUDA-13-compatible environment and recommends a container or checkout-based local install, not the lagging PyPI path. GPU/model/dataset rights, dependency compatibility, infrastructure cost and safe handling of prompts or tool calls require local review. Apache-2.0 is documented; licence/attribution files were not independently audited.","updated_at":"2026-10-17T20:39:39Z","url":"https://research.wusong.net/cases/molt-agentic-rl-framework","sources":[{"label":"Molt: agentic reinforcement-learning framework — README","url":"https://github.com/NVIDIA-NeMo/labs-molt","kind":"code-repository","is_primary":1}]},{"slug":"morsi-world-shock-ml-research-case-study","title":"Morsi portfolio: selected World Shock Lab ML comparison method","summary":"An owner-written portfolio case study describes comparing machine-learning and linear predictions around economic announcements with prespecified tests and explicit negative findings.","resource_type":"use-case","disciplines":["computer-science","economics"],"tasks":["data-analysis","experiment-design"],"evidence_level":"project-documented","outcome_verification":"author-reported","workflow_level":"L2","workflow_note":"Scoped L2 for the selected World Shock Lab case study within the original portfolio. Researchers define timing, preregistered splits, metrics and a decision rule, then compare models, score held-out data and examine uncertainty. This is not a level for every portfolio project or an autonomous trading system.","source_checked_date":"2026-10-19","limitations":"Portfolio README and selected owner-written World Shock Lab static text only. The separate ATLAS page is context, not the classified method; sibling repositories are not silently substituted for the portfolio identity. No release, live forecast, raw data, figure, test or model was audited/executed. The author reports no evidence that ML adds value; this is not independently verified superiority. Causal labels depend on assumptions, with timing, proxy, weak-instrument, small held-out sample and prospective-scoring limits. Future forecasts are not observed performance. No investment advice or trading action is supplied. Static pages/HEAD are moving and non-atomic.","adoption_requirements":"Untested suggestion: examine the selected method, data timing, design, baselines and uncertainty before a separate research trial.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/morsi-world-shock-ml-research-case-study","sources":[{"label":"Original portfolio README","url":"https://raw.githubusercontent.com/shmorsi/shmorsi.github.io/HEAD/README.md","kind":"code-repository","is_primary":1},{"label":"Selected World Shock Lab case study","url":"https://shmorsi.github.io/projects/world-shock-lab.html","kind":"project-documentation","is_primary":1}]},{"slug":"multimodal-manipulator-control-sdl","title":"Multimodal Control of Manipulators: Coupling Kinematics and Vision for Self-Driving Laboratory Operations","summary":"A motion-execution framework for self-driving laboratory integration, combining visual object detection and tracking with manipulator kinematics and path planning.","resource_type":"framework","disciplines":["engineering","robotics"],"tasks":["experiment-execution","image-analysis"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level workflow is unspecified. Researchers set objectives, validate assumptions and outputs, and supervise any adoption.","source_checked_date":"2026-10-17","limitations":"Abstract-only substantive review. Evaluation is described as simulation, not physical laboratory validation; no implementation, quantitative metric values or chemical discovery outcome was verified. This is classical vision/planning support for SDLs, not evidence of an LLM or learned controller.","adoption_requirements":"To investigate adoption, obtain the full methods, compatible manipulator and camera models and a simulation environment; implementation availability is unconfirmed. Physical use requires calibration, collision/safety controls and qualified oversight.","updated_at":"2026-10-17T14:30:44Z","url":"https://research.wusong.net/cases/multimodal-manipulator-control-sdl","sources":[{"label":"Multimodal Control of Manipulators: Coupling Kinematics and Vision for Self-Driving Laboratory Operations","url":"https://www.mdpi.com/2218-6581/15/1/17","kind":"doi","is_primary":1}]},{"slug":"multimodule-health-interview-thematic-analysis-comparison","title":"Multimodule LLM thematic analysis of digital-health interviews: comparative workflow study","summary":"A comparative study separates LLM-assisted health-interview analysis into code extraction, code combination and theme generation in independent conversations. It compares ChatGPT GPT-5.2-thinking, Opus 4.6 and Gemini 3 Pro across three datasets and five workflow strategies, using embedding-based alignment with human-generated themes.","resource_type":"workflow-process","disciplines":["medicine","social-science"],"tasks":["data-analysis","workflow-orchestration"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Researcher-supervised process; automation level intentionally unspecified. No autonomous research or clinical deployment asserted.","source_checked_date":"2026-10-17","limitations":"Partial publisher paper-excerpt review: extraction backend cut the body at about 50,000 characters; figures, supplements, discussion tail and underlying transcripts were not fully reviewed. Authors ran each model-workflow condition once through consumer web interfaces. Semantic similarity and valid code identifiers do not establish qualitative interpretive correctness. No independent execution or reproduction; no universal model-ranking claim. The general interview-method section describes one-hour in-person visits, while the Table 4 footnote describes COPD remote interviews after a 12-month programme. This cohort-context difference is unresolved; no uniform interview-setting claim is made. The cited prior framework has unresolved bibliographic identity and is not assumed to alias another resource.","adoption_requirements":"Approved access to interview data and LLM platforms, qualitative expertise and privacy/consent controls; manual organization between conversations. Authors used paid personal subscriptions, but that is a reported study configuration, not a recommendation to upload sensitive data to consumer accounts.","updated_at":"2026-10-17T18:33:50Z","url":"https://research.wusong.net/cases/multimodule-health-interview-thematic-analysis-comparison","sources":[{"label":"Multimodule Human–Artificial Intelligence Collaboration Pipeline for Large Language Model–Assisted Thematic Analysis Across Digital Health Interview Studies: Comparative Evaluation Study","url":"https://doi.org/10.2196/96129","kind":"doi","is_primary":1}]},{"slug":"muse-prompts-science-communication-reference","title":"Muse image prompt collection: science-communication example","summary":"A broader creative prompt collection includes a specific source-research-to-science-infographic example concerning the giant-impact hypothesis.","resource_type":"tutorial-template","disciplines":["arts","education","physics"],"tasks":["knowledge-synthesis","scientific-writing"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level assigned: the selected science-communication prompt requests source research and visual synthesis, but researcher checking/approval and actual tool execution are not established. This is a reference/template, not validated scientific accuracy or vendor capability evidence.","source_checked_date":"2026-10-19","limitations":"Selected README introduction/tool-use examples, especially lines138–146, read; not all987lines, images, external posts or vendor documentation. The curator calls examples official/source-verified and attributes capabilities to Meta Muse; those are unverified curator claims, not primary proof of vendor origin, availability or effectiveness. No generated visual, astronomical factual claim, citation accuracy or self-refinement result evaluated. Scope excludes the wider consumer/marketing catalogue. Null workflow level is reasoned, not a generic template exemption. AI editorial source review only; no installation, source code execution, provider/model calls or outcome reproduction.","adoption_requirements":"Read the documented prompt as a reference. Access, provider terms, data handling, costs and current model capability were not verified; no service connected or prompt submitted.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/muse-prompts-science-communication-reference","sources":[{"label":"imagineVid/Awesome-muse-image-prompts-and-skills primary-readme","url":"https://raw.githubusercontent.com/imagineVid/Awesome-muse-image-prompts-and-skills/356fdf034430f11f93a8bf2f46731725b1ce2785/README.md","kind":"code-repository","is_primary":1}]},{"slug":"neural-3d-reconstruction-laboratory-robots-benchmark","title":"Neural 3D reconstruction benchmark for laboratory robots","summary":"A compute-platform benchmark examines neural 3D reconstruction trade-offs for laboratory robots, comparing NeRF, Gaussian Splatting and a preliminary SAM3D assessment.","resource_type":"benchmark-dataset","disciplines":["computer-science","robotics"],"tasks":["data-analysis","image-analysis"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Platform-level automation is author-described; workflow level left unspecified because scope and researcher responsibilities depend on the implementation. Human safety and research oversight remain required.","source_checked_date":"2026-10-17","limitations":"Abstract-only review. Results are author-reported; device specifications, datasets, metric definitions and implementation settings were not inspected. SAM3D assessment is explicitly preliminary. Workshop peer review is asserted in arXiv comments, not independently confirmed; evidence is conservatively labelled preprint. No dataset distribution is implied by the benchmark-dataset taxonomy.","adoption_requirements":"Editorial, untested guidance: consult the paper for exact hardware, scene data and metrics; evaluate on the intended robot compute stack and validate geometric errors under supervised, safe manipulation tests. No downloadable benchmark package was verified.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/neural-3d-reconstruction-laboratory-robots-benchmark","sources":[{"label":"Cross-Platform Benchmark of Neural 3D Reconstruction for Autonomous Laboratory Robots","url":"https://arxiv.org/abs/2608.26383","kind":"arxiv","is_primary":1}]},{"slug":"neurocortex-srdf","title":"NeuroCortex SRDF","summary":"NeuroCortex SRDF is a repository-documented research framework for controlled runtime adaptation in machine-learning systems. Its design separates candidate generation, evaluation and authorization, with traceable decisions and retained prior state when an adaptation is rejected.","resource_type":"framework","disciplines":["computer-science"],"tasks":["agent-evaluation","experiment-design","optimization"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation is unspecified. Researchers choose goals, inputs, permissions and evaluation criteria and remain responsible for interpreting outputs.","source_checked_date":"2026-10-17","limitations":"Partial README documentation review only; code, archived protocols, data and result artifacts were not independently inspected or executed. Architecture is conceptual and components may differ across experiments. README empirical results are author-reported, not reproduced; no universal self-improvement, general intelligence or generalization claim is accepted. The README itself notes a missing original S04 selection notebook and distinguishes folds from independent datasets.","adoption_requirements":"Select a specific protocol and matching source revision; inspect pyproject.toml, requirements and experiment artifacts before setting up a Python environment. Respect dataset rights, avoid sensitive data in shared notebooks, and record seeds, preprocessing and statistical units. README identifies Apache-2.0 but the licence file and third-party terms were not audited.","updated_at":"2026-10-17T20:39:39Z","url":"https://research.wusong.net/cases/neurocortex-srdf","sources":[{"label":"NeuroCortex SRDF — README","url":"https://github.com/mohammedqaidalathwary/neurocortex-srdf","kind":"code-repository","is_primary":1}]},{"slug":"nlp05-agentic-literature-review","title":"Agentic AI literature-review coursework pipeline (NLP-05)","summary":"A documented coursework pipeline combines scholarly PDF collection/parsing, structured extraction and role-based literature synthesis, comparison and report preparation.","resource_type":"framework","disciplines":["computer-science","multidisciplinary"],"tasks":["data-extraction","knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Intended L2 for fixed collection/parsing/schema-extraction and role-based synthesis/comparison/report stages. User configuration and CLI stages do not establish adaptive L3/L4. Similarity matching is not a guarantee of zero hallucination or page-level provenance; final researcher checks are not documented.","source_checked_date":"2026-10-19","limitations":"Full README only; code, demos, PDFs, GROBID service, claimed institution/mentors, licence file and model availability not independently verified. Source says zero-hallucination/page-provenance guarantee from vector similarity; no such guarantee accepted. Fixed role orchestration does not prove planning/iteration. Repository tree and feature claims may differ; no completion/production-readiness assumption. Licence, release status and effectiveness remain unverified. AI editorial review only; no source code, models or installation executed, no production/status/publication changes.","adoption_requirements":"README names Python3.10+, optional Gemini extraction and optional GROBID with parser fallback; inspect current dependencies, paper rights and provider/data handling before any separately authorised trial. No services or downloads started.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/nlp05-agentic-literature-review","sources":[{"label":"Chaitanya-G41/Agentic-AI-for-literature-review primary-readme","url":"https://raw.githubusercontent.com/Chaitanya-G41/Agentic-AI-for-literature-review/HEAD/README.md","kind":"code-repository","is_primary":1}]},{"slug":"numguard","title":"numguard: numerical edge-case testing with an optional LLM repair agent","summary":"Repository-documented testing tool for numerical array functions, combining edge-case inputs, references and metamorphic checks with an optional experimental LLM that suggests specifications or function rewrites.","resource_type":"tool","disciplines":["computer-science","mathematics"],"tasks":["research-coding"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level workflow is unspecified. Researchers set objectives, validate assumptions and outputs, and supervise any adoption.","source_checked_date":"2026-10-17","limitations":"Partial repository documentation review only; no installation, security audit or outcome reproduction. The experimental agent success rate is unmeasured. Finite cases do not prove correctness; permissive specifications can hide bugs, and real optional backends are not exercised in the stated CI tests.","adoption_requirements":"Python 3.10+, NumPy and click; suitable functions, reference implementations and researcher-reviewed specifications. Optional model access entails API costs and sending code to a provider: review confidentiality and institutional policy first.","updated_at":"2026-10-17T14:30:44Z","url":"https://research.wusong.net/cases/numguard","sources":[{"label":"GitHub - bhaskargurram-ai/numguard: Numerical edge-case testing for array code. Runs your functions on NaN, ±inf, overflow, subnormals, cancellation and float16/32/64, and compares them with high-precision references and metamorphic properties. Works as a pytest plugin, a CLI and a GitHub Action. Optional LLM agent suggests fixes that must pass every check. pip install numguard-ci · GitHub","url":"https://github.com/bhaskargurram-ai/numguard","kind":"code-repository","is_primary":1}]},{"slug":"obsidian-academic-archive-rules","title":"Academic Obsidian archiving rules for Codex/Claude Code","summary":"A documented single-file academic archiving workflow connects selected papers and supplements to source-traceable Obsidian notes, version comparisons and research-method links.","resource_type":"workflow-process","disciplines":["multidisciplinary"],"tasks":["data-management","knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Intended L2 for a fixed six-stage academic archive process. Researchers choose papers/questions/scope and confirm concrete plans; reading, identity/version checks, archiving, evidence-bounded linking and final review remain source-prescribed rather than verified enforcement.","source_checked_date":"2026-10-19","limitations":"Full Chinese README only; core llm-wiki-starter.md, images, agent tools and actual vault not executed/audited. Reviewed as a rule/reference workflow, not an installed plugin or automatic task. Human confirmation/source rechecking are prescribed, not observed. README local storage does not establish no external model egress; model/provider choices and permissions require separate review. Reference to Karpathy LLM Wiki is acknowledged lineage, not same resource/canonical substitution. AI editorial review only; no source code, models or installation executed, no production/status/publication changes.","adoption_requirements":"An Obsidian-style folder plus a file-reading/writing agent environment are described. Read original rules/permissions and choose public/synthetic material before any separate trial; no file-agent invoked.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/obsidian-academic-archive-rules","sources":[{"label":"NoNightWatch/Obsidian-Codex-CC-for-academy primary-readme","url":"https://raw.githubusercontent.com/NoNightWatch/Obsidian-Codex-CC-for-academy/HEAD/README.md","kind":"code-repository","is_primary":1}]},{"slug":"onhand-research-paper-reader","title":"Onhand: contextual research-paper reader","summary":"Onhand documents a Chromium side-panel AI reading assistant for research papers, webpages and PDFs. Its official guide describes passage highlighting, paper explanations, following references and comparing clearly related tabs; the runtime is browser-hosted, while model inference uses configured external services.","resource_type":"tool","disciplines":["multidisciplinary"],"tasks":["knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L1","workflow_note":"L1 applies only to the user-invoked explanation of a passage in an already open research paper or PDF, with highlighted supporting text that the reader can check in context. It is not a universal classification of optional citation-following, multi-tab comparison or other browser-tool modes. Users select material and requests; output checking and approved provider/data exposure remain necessary research safeguards, not proof that any actual user checked an answer. Provider/related-tab transmission, powerful permissions and restore limitations remain disclosed; no extension, provider or privacy traffic test was performed.","source_checked_date":"2026-10-17","limitations":"Documentation-only review of the README, official paper-reading guide, repository privacy policy and LICENSE; no extension installation, browser run, provider call or source-code audit. Highlighted passages and cited answers are documented capabilities, not guarantees of answer truth, citation accuracy or complete paper coverage. Local session storage does not mean local inference: relevant page/PDF content, prompts, selected text and possible screenshots may leave the device, and related-tab titles/URLs and relevant content can be inspected. The policy describes direct provider calls or an Onhand Free Cloudflare Worker forwarding to OpenAI, with required free-tier diagnostics; these privacy claims were not independently traffic-audited. No universal full-document upload or no-upload guarantee is established. Powerful debugging/scripting permissions require trust. README says changed-page/missing-tab restore remains in progress; archived runtime plans and historical evaluations are not current capability guarantees. No research accuracy, performance or productivity outcome evaluated.","adoption_requirements":"Requires a compatible Chromium extension environment and explicit acceptance of powerful browser debugging/scripting permissions. The README documents free hosted usage with daily caps and required diagnostics, Codex sign-in, or supported provider API keys; provider access and terms vary. Review active and related tabs before use, close sensitive material, and obtain authorization before any confidential, licensed or identifying text/image is sent to providers or the hosted Worker. Local saved sessions can retain sensitive artifacts and credentials; review deletion and retention settings.","updated_at":"2026-10-17T22:18:09Z","url":"https://research.wusong.net/cases/onhand-research-paper-reader","sources":[{"label":"Onhand — primary repository README","url":"https://github.com/Phineas1500/Onhand","kind":"code-repository","is_primary":1},{"label":"Onhand — repository privacy policy","url":"https://raw.githubusercontent.com/Phineas1500/Onhand/HEAD/website/privacy.html","kind":"project-documentation","is_primary":1},{"label":"Onhand — official research-paper reader guide","url":"https://useonhand.com/ai-research-paper-reader.html","kind":"project-documentation","is_primary":1},{"label":"Onhand — repository Apache-2.0 LICENSE","url":"https://raw.githubusercontent.com/Phineas1500/Onhand/HEAD/LICENSE","kind":"project-documentation","is_primary":1}]},{"slug":"ontological-mediation-explainable-ai","title":"Ontological Mediation for Explainable AI","summary":"A conceptual framework for assessing AI-generated explanations by separating ontological clarification, ontological mediation and justificatory assessment. The repository companion documents how model-derived explanatory claims can be examined against explicit evidential, contextual and normative criteria.","resource_type":"framework","disciplines":["computer-science","humanities"],"tasks":["data-analysis"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No fixed automation level: a conceptual framework for researcher assessment of AI explanations. Human domain judgment remains necessary; no autonomous justificatory assessor was verified.","source_checked_date":"2026-10-17","limitations":"Reviewed repository README and contribution document only, not the SciELO manuscript, formal proofs, code or case-study data. The source explicitly states that operational mechanisms, formal properties, computational realization and empirical effectiveness need further specification and evaluation. The README says an accepted workshop manuscript was withdrawn and absent from proceedings; no peer-reviewed-publication or machine-verified label is asserted. Post-preprint CRB generalizations are hypotheses, not established preprint results.","adoption_requirements":"Requires domain expertise, explicit ontological commitments, access to model-derived concepts or rules, and a defensible evidential/normative context. This is not a turnkey tool or clinical decision system. Preserve privacy and data-use rights when considering sensitive explanatory examples.","updated_at":"2026-10-17T21:43:09Z","url":"https://research.wusong.net/cases/ontological-mediation-explainable-ai","sources":[{"label":"Ontological Mediation for Explainable AI — repository README","url":"https://github.com/em4ttos/Ontological-Mediation-in-Explainable-Artificial-Intelligence","kind":"code-repository","is_primary":1},{"label":"Ontological Mediation for Explainable AI — targeted primary documentation","url":"https://raw.githubusercontent.com/em4ttos/Ontological-Mediation-in-Explainable-Artificial-Intelligence/HEAD/docs/research-contributions/from-ontological-unpacking-to-justificatory-assessment.md","kind":"project-documentation","is_primary":1}]},{"slug":"ontology-based-drug-repurposing-infrastructure","title":"Ontology-based drug-repurposing infrastructure and prediction methods","summary":"An arXiv abstract documents ontology-labelled compound-target data infrastructure and two computational association-prediction approaches for drug-repurposing research.","resource_type":"framework","disciplines":["biomedicine","computer-science"],"tasks":["data-analysis","drug-repurposing","knowledge-synthesis"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level assigned. The abstract describes ontology-labelled compound–target associations and information-retrieval versus matrix-factorisation/similarity prediction methods. Human orchestration, output validation, operational links and stopping conditions are not documented.","source_checked_date":"2026-10-19","limitations":"Official arXiv abstract only. No full paper, repository, service, original data or evaluation protocol was audited. Although the authors report studying both tools, outcome values and comparison validity were not assessed. This is the original 2018 preprint; present-day tool availability is unknown. No drug recommendation, clinical effectiveness or medical inference is supplied.","adoption_requirements":"Untested suggestion: inspect the original paper and establish method, data, rights and access requirements before any research trial. Clinical validity is not established.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/ontology-based-drug-repurposing-infrastructure","sources":[{"label":"Primary preprint abstract","url":"https://arxiv.org/abs/1807.09754","kind":"arxiv","is_primary":1}]},{"slug":"openlabs-bounded-research-runtime-reference","title":"OpenLabs: bounded research-runtime reference architecture","summary":"A public reference architecture describes recoverable, resource-bounded agent research attempts across separate domain laboratories and private research state.","resource_type":"framework","disciplines":["computer-science","multidisciplinary"],"tasks":["research-coding","workflow-orchestration"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level assigned. The architecture documents seeded, bounded research attempts, agent-chosen scientific steps, leases, resource controls, evidence gates and recovery. Its independent reviewers are fresh agent/provider sessions; human validation of scientific conclusions is not established. Hashes and a needs-human stop do not establish L3/L4 oversight.","source_checked_date":"2026-10-19","limitations":"README and runtime/constraint documents only. Code, hooks, private campaigns, artefacts, databases, claimed papers/results and services were not inspected or run. The source describes broad runtime permissions; an attempt working directory is a convention, not kernel-enforced write isolation, and resource limits are not security containment. Public code alone does not reproduce private campaign state. Hash/replay checks and fresh agent reviewers do not prove scientific truth or human acceptance. Internal readiness is not submission or authorship approval. No source permission or runtime setting was applied. Moving HEAD documents are unpinned and non-atomic.","adoption_requirements":"Untested architecture reference: establish enforceable access boundaries, required private-state availability and explicit human conclusion acceptance before a trial; do not inherit source runtime permissions.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/openlabs-bounded-research-runtime-reference","sources":[{"label":"Project README","url":"https://raw.githubusercontent.com/chainstart/openlabs/HEAD/README.md","kind":"code-repository","is_primary":1},{"label":"Research runtime documentation","url":"https://raw.githubusercontent.com/chainstart/openlabs/HEAD/docs/autonomous-runtime.zh.md","kind":"project-documentation","is_primary":1},{"label":"Minimal-constraint documentation","url":"https://raw.githubusercontent.com/chainstart/openlabs/HEAD/docs/minimal-constraints.zh.md","kind":"project-documentation","is_primary":1}]},{"slug":"orchestra-ai-research-skills","title":"AI Research Skills (Orchestra Research)","summary":"Orchestra Research provides an AI research skills library covering research-agent tasks and technical research workflows.","resource_type":"agent-skill","disciplines":["computer-science","multidisciplinary"],"tasks":["research-coding","scientific-writing","workflow-orchestration"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Repository documentation is not evidence that agents autonomously produce valid research. Individual skills and dependencies were not audited or run; inspect commands and permissions before use.","adoption_requirements":"A compatible agent environment and technical expertise to review skill instructions and verify generated work.","updated_at":"2026-10-16T14:58:30Z","url":"https://research.wusong.net/cases/orchestra-ai-research-skills","sources":[{"label":"Comprehensive open-source library of AI research and engineering skills ...","url":"https://github.com/orchestra-research/AI-research-SKILLs","kind":"code-repository","is_primary":1}]},{"slug":"panda-electrodeposited-polymer-films","title":"PANDA: electrodeposited polymer-film self-driving laboratory","summary":"PANDA combines automated electrodeposition with optical and electrochemical characterization to study functional polymer films.","resource_type":"framework","disciplines":["chemistry","materials-science"],"tasks":["experiment-design","experiment-execution","optimization"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Platform-level automation is author-described; workflow level left unspecified because scope and researcher responsibilities depend on the implementation. Human safety and research oversight remain required.","source_checked_date":"2026-10-17","limitations":"Abstract-only canonical source review; supporting paper excerpts were inspected separately but software and hardware were not executed. Calibration, repeatability and optimization are author-reported; no independent performance or cost verification. The arXiv page links a journal DOI, but the journal article was not independently reviewed.","adoption_requirements":"Editorial, untested guidance: requires electrochemistry expertise, compatible robotic fluid handling, potentiostat and optical hardware, sample fabrication and laboratory safety controls; not a turn-key software installation. Verify code availability and licence with the authors before reuse.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/panda-electrodeposited-polymer-films","sources":[{"label":"PANDA: A self-driving lab for studying electrodeposited polymer films","url":"https://arxiv.org/abs/2406.17725","kind":"arxiv","is_primary":1}]},{"slug":"paper-litflow","title":"paper-litflow: chapter-organized literature extraction","summary":"Python command-line workflow documented to classify PDFs by manuscript section, extract quotations and export chapter-organized Word material. Its extract-v2 route describes page markers, local quote matching and resumable processing; a companion skill lets an AI assistant drive the workflow.","resource_type":"tool","disciplines":["multidisciplinary"],"tasks":["data-extraction","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation unspecified. The researcher defines the task, controls access and checks sources and outputs.","source_checked_date":"2026-10-17","limitations":"README-only review; no commands, tests, extraction accuracy or performance claims reproduced. Long-text character limits and keyword filtering can omit relevant evidence; OCR can corrupt quotations. A Word export of excerpts is not a validated literature review or exhaustive search. Author timing and cost assertions are not adopted as evaluated outcomes.","adoption_requirements":"Python dependencies, chapter configuration and accessible PDFs are needed; a model endpoint/API is required for AI steps. Tesseract is optional for scanned pages. Confirm rights and privacy before transmitting documents, protect API keys, budget API costs and manually check omissions and context.","updated_at":"2026-10-17T17:46:50Z","url":"https://research.wusong.net/cases/paper-litflow","sources":[{"label":"chanada440824/paper-litflow — primary README","url":"https://github.com/chanada440824/paper-litflow","kind":"code-repository","is_primary":1}]},{"slug":"papermind-ai-pranjali","title":"PaperMind AI: paper-grounded reading and chat","summary":"A documented web tool accepts a research-paper PDF for structured AI analysis and chat constrained to its extracted text.","resource_type":"tool","disciplines":["multidisciplinary"],"tasks":["knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level assigned: paper selection, structured analysis and chat are documented, but researcher output-validation/approval is not established. Documentation is not proof of grounding or correctness.","source_checked_date":"2026-10-19","limitations":"Full README only; no screenshots, code, actual PDF, model endpoint or app trial. Backend-secret and selected-text restrictions are documentation claims, not privacy/security guarantees. Uploaded/extracted text may be handled by configured model/backend; no assurance of local-only data. Source-named model, bounds and dependencies not independently verified. Classification rests on documentation only; no release, evaluation or licence was independently verified. AI editorial review only; no source code, models or installation executed, no production/status/publication changes.","adoption_requirements":"README names React/Node/Express, MongoDB, Gemini configuration and PDF parser; inspect provider/data handling and rights before any separate trial. No setup or credentials accessed.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/papermind-ai-pranjali","sources":[{"label":"Pranjaliingle293/PaperMind-Ai primary-readme","url":"https://raw.githubusercontent.com/Pranjaliingle293/PaperMind-Ai/HEAD/README.md","kind":"code-repository","is_primary":1}]},{"slug":"paperoai-distillai","title":"PaperoAI / DistillAI literature summarization","summary":"Repository-documented pharmaceutical-literature summarization application, called DistillAI in the PaperoAI repository README, using PubMed identifiers or paper PDFs to produce structured summaries.","resource_type":"tool","disciplines":["biomedicine"],"tasks":["data-extraction","knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Workflow level left unspecified for this resource-level listing. Researchers choose scope, inspect inputs and methods, and verify generated outputs before reuse.","source_checked_date":"2026-10-17","limitations":"README-only review. The repository is PaperoAI but the README title is DistillAI and its clone command is a placeholder; use the actual GitLab source and verify setup. Accuracy, throughput, improved feedback and drug-discovery acceleration claims lack reproduced evidence here. Summaries can omit context or hallucinate; PubMed input is described as abstract/metadata, not guaranteed full text. No security or clinical validation was performed.","adoption_requirements":"README lists Node.js v18+, PostgreSQL, Redis and Gemini/NCBI configuration. Confirm paper rights and provider privacy terms before sending text to external services. Audit authentication and mock seeded accounts before any deployment.","updated_at":"2026-10-17T14:30:44Z","url":"https://research.wusong.net/cases/paperoai-distillai","sources":[{"label":"PaperoAI / DistillAI literature summarization — primary repository README","url":"https://gitlab.com/narvin21/paperoai","kind":"code-repository","is_primary":1}]},{"slug":"paperpal","title":"Paperpal","summary":"Paperpal advertises a research-writing workspace combining source uploads, cited drafting assistance, paraphrasing and submission checks.","resource_type":"tool","disciplines":["multidisciplinary"],"tasks":["literature-research","scientific-writing"],"evidence_level":"vendor-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Stored homepage has omissions and marketing claims. Citation correctness, integrity-check accuracy, privacy and publication benefits were not established; AI detection is not proof of authorship.","adoption_requirements":"Service access, permitted source/manuscript files, and researcher review of citations and substantive claims.","updated_at":"2026-10-16T14:58:30Z","url":"https://research.wusong.net/cases/paperpal","sources":[{"label":"Paperpal: AI Academic Writing Tool - Comprehensive AI Research ...","url":"https://paperpal.com","kind":"vendor-documentation","is_primary":1}]},{"slug":"paperqa2","title":"PaperQA2: language agents achieving superhuman synthesis of scientific knowledge","summary":"A frontier language-model agent optimized for factuality that performs literature research tasks over scientific papers. The authors report it matches or exceeds subject-matter-expert performance on three realistic literature research tasks, and that its cited Wikipedia-style summaries are more accurate than existing human-written Wikipedia articles.","resource_type":"use-case","disciplines":["biology","multidisciplinary"],"tasks":["knowledge-synthesis","literature-research"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":"L3","workflow_note":"A literature agent coordinates multi-step retrieval and synthesis over scientific papers. Humans define questions, provide corpus access and must check citations and conclusions.","source_checked_date":"2026-10-15","limitations":"Preprint: not peer-reviewed at source-check time. The human-expert comparisons and contradiction validation were run by the authors themselves. Contradiction detection is demonstrated on biology papers; generality to other domains is untested here.","adoption_requirements":"Model API access and a literature corpus (e.g. PDF collections or licensed databases); willingness to spot-check citations, since the core claim is improved factuality on cited summaries, not infallibility.","updated_at":"2026-10-15T17:25:09Z","url":"https://research.wusong.net/cases/paperqa2","sources":[{"label":"Skarlinski et al., Language agents achieve superhuman synthesis of scientific knowledge, arXiv:2409.13740 (preprint, 2024)","url":"https://arxiv.org/abs/2409.13740","kind":"arxiv","is_primary":1},{"label":"arXiv DOI record","url":"https://doi.org/10.48550/arXiv.2409.13740","kind":"doi","is_primary":0},{"label":"Code repository (PaperQA2)","url":"https://github.com/Future-House/paper-qa","kind":"code-repository","is_primary":0}]},{"slug":"papyrus-research-assistant","title":"Papyrus: embedding-backed research-paper corpus search","summary":"Source-documented research assistant with a Rust CLI for ingesting paper lists or local PDF/Markdown collections and querying the corpus through embeddings, alongside a Qt/C++ paper-manager GUI.","resource_type":"tool","disciplines":["multidisciplinary"],"tasks":["data-management","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation unspecified. Researchers select the corpus and embedding configuration, control data transfer, and verify retrieved material and any subsequent claims.","source_checked_date":"2026-10-17","limitations":"Partial source/documentation review, not a build, security audit or runtime test. GUI README mixes a Qt5 build heading with Qt6 dependencies; setup may need maintainer clarification. No retrieval-quality benchmark verified. Audit document rights and embedding endpoint privacy before use; repository tree also includes a Sci-Hub bridge, which was not exercised or endorsed.","adoption_requirements":"Source build and dependencies, configured embedding service/model, document collection and storage. Reviewed code references Ollama; complete installation compatibility was not established. Check document rights and endpoint privacy before ingestion.","updated_at":"2026-10-17T14:30:44Z","url":"https://research.wusong.net/cases/papyrus-research-assistant","sources":[{"label":"Papyrus repository","url":"https://gitlab.com/marcusbritanicus/papyrus","kind":"code-repository","is_primary":1},{"label":"Papyrus CLI source and command documentation","url":"https://gitlab.com/api/v4/projects/9529742/repository/files/crates%2Fpapyrus-cli%2Fsrc%2Fmain.rs/raw?ref=main","kind":"code-repository","is_primary":1},{"label":"Papyrus GUI README","url":"https://gitlab.com/api/v4/projects/9529742/repository/files/crates%2Fpapyrus-gui%2FREADME.md/raw?ref=main","kind":"code-repository","is_primary":1}]},{"slug":"particle-swarm-lab-manipulator-control-study","title":"Particle-swarm tuning for laboratory manipulator control research","summary":"A study uses particle-swarm optimisation within research comparing mobile laboratory-manipulator controllers.","resource_type":"use-case","disciplines":["engineering","robotics"],"tasks":["experiment-design","optimization"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":"L2","workflow_note":"Scoped L2 for a researcher-designed controller comparison: particle-swarm optimisation tunes control parameters, followed by simulation and laboratory comparison. PSO is the supported computational-intelligence role; the sliding-mode controller is not an LLM agent. No L3/L4 or physical-safety certification is established.","source_checked_date":"2026-10-19","limitations":"Official abstract and selected HTML method/tuning passages only. Equations, figures, complete results, robot, code and underlying data were not audited or executed. Related-work neural controllers are not this paper’s implementation. Claimed gentleness and control advantages remain author-reported; they do not validate hazardous-chemical handling or deployment safety. No physical action or experiment was performed.","adoption_requirements":"Untested suggestion: inspect the complete method and establish independent simulation, engineering/safety review and data checks before any trial.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/particle-swarm-lab-manipulator-control-study","sources":[{"label":"Selected primary method and PSO passages","url":"https://arxiv.org/html/2602.06977","kind":"arxiv","is_primary":1}]},{"slug":"pi-academic-writer","title":"Academic Writer for Pi and oh-my-pi","summary":"Repository-documented academic writing package for humanities researchers, using source-linked evidence notes, approved outlines, section drafting and review.","resource_type":"agent-skill","disciplines":["humanities"],"tasks":["literature-research","scientific-writing"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation is unspecified; researchers choose sources, approve interpretations and remain responsible for results.","source_checked_date":"2026-10-17","limitations":"README-only review, no source-code audit or execution. README headline assurances about citation verification do not establish factual correctness: the same README states that citation audit does not prove support. Original-source checking, authorship policies and model/provider privacy review remain necessary. Queue metadata says Claude Code plugin; current README describes the Pi/oh-my-pi port.","adoption_requirements":"Pi with its documented companion packages or oh-my-pi; access to a configured model and permitted source materials. Pandoc is needed for document export; PDF additionally needs a supported TeX engine. Zotero is optional.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/pi-academic-writer","sources":[{"label":"JackMatanky/pi-academic-writer — primary README","url":"https://github.com/JackMatanky/pi-academic-writer","kind":"code-repository","is_primary":1}]},{"slug":"polymer-agent","title":"Polymer-Agent","summary":"Terminal-integrated closed-loop polymer structure-property predictor using LLM reasoning for property prediction, property-guided structure generation and structure modification.[7]","resource_type":"tool","disciplines":["chemistry","materials-science"],"tasks":["data-analysis","materials-discovery"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans set objectives, supply data and check outputs.","source_checked_date":"2026-10-16","limitations":"Author-reported capabilities; not executed or benchmarked in this review. Code availability not independently verified; the PMC deposit was noted but full text not inspected.","adoption_requirements":"Not verified in checked records; assess the article's code/data statement for software access before adoption.","updated_at":"2026-10-17T08:00:32Z","url":"https://research.wusong.net/cases/polymer-agent","sources":[{"label":"Polymer-Agent: Large Language Model Agent for Polymer Design - Journal of Chemical Information and Modeling (ACS)","url":"https://doi.org/10.1021/acs.jcim.6c00343","kind":"doi","is_primary":1}]},{"slug":"polymer-ml-model-development-process","title":"Polymer ML model development: representation, cleaning and design workflow","summary":"A review-derived process reference linking polymer representation, data preparation, property prediction and materials design, including inverse design, virtual screening and multi-objective optimization.","resource_type":"workflow-process","disciplines":["materials-science"],"tasks":["data-analysis","data-management","materials-discovery"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Workflow level intentionally unspecified for a reusable component or conceptual reference; researcher oversight and task-specific validation remain necessary.","source_checked_date":"2026-10-17","limitations":"Abstract-only. This is a conceptual, review-derived process reference, not an implemented model or verified high-reliability pipeline. Specific databases, descriptors, validation splits, code, design results and quantitative gains were not read. The arXiv HTML route returned404.","adoption_requirements":"Editorial, untested requirements: polymer data with appropriate rights, domain-informed representation choices and a validation plan. Exact databases, algorithms, compute requirements and experimental protocols must be obtained from the complete article.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/polymer-ml-model-development-process","sources":[{"label":"Applications of Machine Learning in Polymer Materials: Property Prediction, Material Design, and Systematic Processes","url":"https://arxiv.org/abs/2510.26100","kind":"arxiv","is_primary":1}]},{"slug":"priorwork","title":"Priorwork: agent-assisted social-science literature reviews","summary":"Repository-documented toolkit for social-science literature reviews using an external AI agent, with CLI and VS Code interfaces. It documents literature searching, screening with reasons, citation chasing, evidence cards and report export. ","resource_type":"tool","disciplines":["social-science"],"tasks":["knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L3","workflow_note":"L3 applies to the README-documented standard-mode literature survey performed by an external AI agent using Priorwork state and CLI tools, not to the extension alone or every operating mode. The agent connects scope, scholarly search, screening with reasons, citation chasing, evidence cards, checks and export. Standard mode pauses for researcher scope approval and before cards; researchers can inspect and override decisions. This is documentation-supported workflow scope, not a claim of exhaustive retrieval, fabrication-proof checks, research correctness or independently reproduced outcomes.","source_checked_date":"2026-10-17","limitations":"README-only review; no software run or research outcomes reproduced. Citation checks are pattern-based, not proof of claim truth. Database coverage excludes some domestic journals; SSCI status may be guessed without the licensed list. Abstract-only cards and unavailable full texts limit inference. Keep abstracts and manuscripts private, protect keys and respect text/database rights and the external agent provider policy.","adoption_requirements":"Python 3.10+ or the extension-managed Python environment; an external AI agent and its terms/costs; Git for the documented private-workspace workflow. Semantic Scholar API key recommended; Zotero and licensed SSCI data optional.","updated_at":"2026-10-17T21:22:00Z","url":"https://research.wusong.net/cases/priorwork","sources":[{"label":"yoshida-kd/priorwork primary documentation","url":"https://github.com/yoshida-kd/priorwork","kind":"code-repository","is_primary":1}]},{"slug":"prithvi-eo-2-0","title":"Prithvi-EO 2.0","summary":"Second-generation Earth-observation foundation model jointly developed by IBM, NASA and Jülich Supercomputing Centre for remote-sensing tasks.[9]","resource_type":"tool","disciplines":["earth-environmental"],"tasks":["data-analysis","image-analysis"],"evidence_level":"project-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Model-card description does not establish performance for every sensor, region or task; fine-tuning data requirements apply. Documentation review only - no software execution or independent performance validation.","adoption_requirements":"Compatible multispectral inputs; downstream task data; ML environment, often fine-tuning.","updated_at":"2026-10-17T06:33:49Z","url":"https://research.wusong.net/cases/prithvi-eo-2-0","sources":[{"label":"Prithvi-EO-2.0 - model card (ibm-nasa-geospatial)","url":"https://huggingface.co/ibm-nasa-geospatial/Prithvi-EO-2.0-300M","kind":"project-documentation","is_primary":1}]},{"slug":"pysr","title":"PySR (including official usage skill)","summary":"PySR performs symbolic regression to search for interpretable mathematical expressions from data and documents an optional agent skill.","resource_type":"tool","disciplines":["mathematics","multidisciplinary"],"tasks":["data-analysis","research-coding"],"evidence_level":"project-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"The page recommends low-dimensional data; fitted expressions are not established scientific laws without validation. The agent skill was not executed or audited.","adoption_requirements":"Python and Julia dependencies, a suitable dataset, candidate operators and held-out validation criteria.","updated_at":"2026-10-16T14:58:30Z","url":"https://research.wusong.net/cases/pysr","sources":[{"label":"PySR","url":"https://ai.damtp.cam.ac.uk/pysr","kind":"project-documentation","is_primary":1}]},{"slug":"qualilens","title":"QualiLens","summary":"Local web application documenting researcher-checkpointed LLM assistance for qualitative coding and literature synthesis.","resource_type":"tool","disciplines":["multidisciplinary","social-science"],"tasks":["data-analysis","knowledge-synthesis","transcription"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Defined wizard/checkpoint workflow: researchers set questions and methods and edit codes/themes before continuation; these are documented design features, not verified execution.","source_checked_date":"2026-10-17","limitations":"Local storage does not mean offline inference: data goes directly to the selected LLM provider and synced folders expose it to the sync service. README says no multiple-human intercoder reliability modeling, no Whisper speaker diarization and no Intel Mac support. Tolerant/partial quote matching is not full semantic validation; security and spend protections were not audited.","adoption_requirements":"Python 3.11+ on Apple-silicon Mac or Linux/WSL, supported provider API credentials, and ffmpeg for video; audio transcription needs an OpenAI key.","updated_at":"2026-10-17T11:14:58Z","url":"https://research.wusong.net/cases/qualilens","sources":[{"label":"QualiLens — primary repository","url":"https://github.com/iconerr/qualilens","kind":"code-repository","is_primary":1}]},{"slug":"quallmer","title":"quallmer","summary":"R toolbox for codebook-based LLM-assisted qualitative coding, comparison against human coding, and audit trails.","resource_type":"tool","disciplines":["social-science"],"tasks":["data-analysis"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Defined coding workflow; researchers select codebooks, data, models and gold standards, and interpret discrepancies.","source_checked_date":"2026-10-17","limitations":"No coding accuracy or reliability gains reproduced. API privacy/cost depend on provider; statistical agreement does not itself establish interpretive validity. README uniqueness and ease-of-use claims were not independently checked.","adoption_requirements":"R plus package dependencies and an ellmer-supported model/provider; configure API authentication or a local Ollama model.","updated_at":"2026-10-17T11:14:58Z","url":"https://research.wusong.net/cases/quallmer","sources":[{"label":"quallmer — primary repository","url":"https://github.com/quallmer/quallmer","kind":"code-repository","is_primary":1}]},{"slug":"quantifying-the-ease-of-scientific-discovery","title":"Quantifying the Ease of Scientific Discovery","summary":"Peer-reviewed study proposing a quantitative method for examining the ease of scientific progress, applied to historical discovery timelines for species, chemical elements and minor planets.","resource_type":"workflow-process","disciplines":["multidisciplinary"],"tasks":["data-analysis"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans set objectives, supply data and check outputs.","source_checked_date":"2026-10-17","limitations":"Methodological study; the modelling assumption is the author's argument and was not independently assessed. Only the abstract and metadata of the published version were checked; the full text was not inspected in this review.","adoption_requirements":"Research paper; no software release confirmed in checked sources. Reuse would require the underlying historical discovery datasets.","updated_at":"2026-10-17T09:28:23Z","url":"https://research.wusong.net/cases/quantifying-the-ease-of-scientific-discovery","sources":[{"label":"Quantifying the Ease of Scientific Discovery - Scientometrics (Springer), DOI 10.1007/s11192-010-0232-6","url":"https://doi.org/10.1007/s11192-010-0232-6","kind":"doi","is_primary":1}]},{"slug":"radiology-copilot","title":"Radiology Copilot","summary":"Research and teaching prototype connecting brain MRI segmentation, measured volumes, LLM-assisted report drafts and model-error feedback in one case-centered workspace.","resource_type":"tool","disciplines":["education","medicine"],"tasks":["data-analysis","image-analysis"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Workflow level left unspecified for this resource-level listing. Researchers choose scope, inspect inputs and methods, and verify generated outputs before reuse.","source_checked_date":"2026-10-17","limitations":"README-only review; no code or clinical validation performed. The authors explicitly prohibit clinical decision use. Their small-sample Dice and model-comparison claims are not independently reproduced and are not adopted as effectiveness evidence here. User corrections are not necessarily clinician-reviewed; model generalization and privacy require separate assessment.","adoption_requirements":"Python environment, project dependencies, separately downloaded MONAI weights and appropriately licensed MRI data. LLM tool calling additionally requires a supported provider or local model; without it the documented fallback is deterministic. Confirm rights and consent before any third-party data transfer.","updated_at":"2026-10-17T14:30:44Z","url":"https://research.wusong.net/cases/radiology-copilot","sources":[{"label":"Radiology Copilot — primary repository README","url":"https://github.com/ssrs-cjc/radiology-copilot","kind":"code-repository","is_primary":1}]},{"slug":"rayyan","title":"Rayyan","summary":"Rayyan supports collaborative review screening, deduplication and relevance ranking learned from reviewer decisions.","resource_type":"tool","disciplines":["multidisciplinary"],"tasks":["data-extraction","literature-research"],"evidence_level":"vendor-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"L2 describes the documented human-supervised sequence, not autonomous research. Humans supply inputs, check intermediate results and approve conclusions.","source_checked_date":"2026-10-15","limitations":"Feature guide only; no screening sensitivity or review-quality outcomes checked. Some advanced functions depend on plan or institutional contract. Humans must own eligibility decisions and resolve disagreements.","adoption_requirements":"Rayyan access, exported reference records, eligibility criteria and an agreed team coding scheme.","updated_at":"2026-10-16T14:58:30Z","url":"https://research.wusong.net/cases/rayyan","sources":[{"label":"Getting Started with Rayyan","url":"https://help.rayyan.ai/hc/en-us/articles/22697630697617-Getting-Started-with-Rayyan","kind":"vendor-documentation","is_primary":1}]},{"slug":"reference-error-detection-llm","title":"Detecting reference errors in scientific literature with large language models","summary":"The authors evaluate OpenAI GPT-family language models for reference-error detection using an expert-annotated dataset of statement-reference pairs from journal articles.","resource_type":"use-case","disciplines":["computer-science","multidisciplinary"],"tasks":["automated-review"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":"L1","workflow_note":"L1 task assistance for reference checking, not autonomous scientific judgment. Human adjudication of errors is recommended editorial oversight.","source_checked_date":"2026-10-17","limitations":"Abstract-only evidence; the dataset's size, licensing, inter-annotator agreement and the models' error profiles are unverified, and no tool was run.","adoption_requirements":"Reimplementation would require suitable statement-reference examples and model access; public dataset availability, prompts, cost and licensing were not verified.","updated_at":"2026-10-17T14:17:23Z","url":"https://research.wusong.net/cases/reference-error-detection-llm","sources":[{"label":"https://arxiv.org/abs/2411.06101","url":"https://arxiv.org/abs/2411.06101","kind":"arxiv","is_primary":1}]},{"slug":"research-grade-skills-literature-guide","title":"Research-Grade Skills: curated literature-review guidance","summary":"A separately maintained, K-Dense-derived skill collection includes a structured literature-review guide and explicit documentation of its structural checks and open limitations.","resource_type":"agent-skill","disciplines":["multidisciplinary"],"tasks":["knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Intended L2 for the selected literature-review skill: researcher questions and criteria guide search, screening, extraction/appraisal, synthesis, reference checks and review. This is a separately maintained K-Dense-derived curation, not an independently invented method or a level for the whole collection.","source_checked_date":"2026-10-19","limitations":"README, selected literature-review skill and maintainer advisory review only; not all skills, scripts, CI runs or references were audited. Shared K-Dense lineage also overlaps the separately catalogued Kalaris guidance. Structural syntax/link checks and maintainer self-review do not establish scientific quality or safety. Prestige, citation-count and reputation filters can bias inclusion. The advisory review reports little script testing, optional untrusted free proxies and paid figure-model dependencies. Figures are optional; study counts must not be invented. Moving HEAD documents are unpinned and non-atomic.","adoption_requirements":"Untested suggestion: select only needed instructions, check upstream attribution and per-skill terms, remove unjustified prestige filters, and assess scripts, providers and source validity before use.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/research-grade-skills-literature-guide","sources":[{"label":"Collection README","url":"https://raw.githubusercontent.com/Yuuqq/research-grade-skills/HEAD/README.md","kind":"code-repository","is_primary":1},{"label":"Selected literature-review skill","url":"https://raw.githubusercontent.com/Yuuqq/research-grade-skills/HEAD/scientific-skills/literature-review/SKILL.md","kind":"project-documentation","is_primary":1},{"label":"Maintainer advisory review","url":"https://raw.githubusercontent.com/Yuuqq/research-grade-skills/HEAD/REVIEW.md","kind":"project-documentation","is_primary":1}]},{"slug":"researchrabbit","title":"ResearchRabbit","summary":"ResearchRabbit supports literature discovery from seed papers, reference-library connections and adaptive recommendations.","resource_type":"tool","disciplines":["multidisciplinary"],"tasks":["literature-research"],"evidence_level":"vendor-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Vendor homepage and testimonials do not establish search completeness or improved review quality. Citation-graph exploration supplements rather than guarantees exhaustive searching.","adoption_requirements":"A seed paper or reference library, service access and a defined research question.","updated_at":"2026-10-16T14:58:30Z","url":"https://research.wusong.net/cases/researchrabbit","sources":[{"label":"ResearchRabbit: AI Tool for Smarter, Faster Literature Reviews","url":"https://www.researchrabbit.ai","kind":"vendor-documentation","is_primary":1}]},{"slug":"researchzosho","title":"ResearchZosho","summary":"A documented research assistant plans questions, searches web/scholarly or supplied-document sources, writes source-linked reports and keeps claims for user acceptance or dispute.","resource_type":"tool","disciplines":["humanities","multidisciplinary"],"tasks":["archival-research","knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L3","workflow_note":"Scoped documented L3: the user sets a research question and permitted duration; the system plans and retrieves sources, writes reports and checks citations. Users accept, dispute or retire claims. These are documented capabilities, not verified implementation, universal L4 or demonstrated autonomous science.","source_checked_date":"2026-10-19","limitations":"Commit-pinned README documentation only; repository code, linked guides, figures, dependencies and claimed tests were not executed/audited. No independent scientific or engineering result verified. Source describes planning, source retrieval/checks and human acceptance as capabilities; citation completeness/source independence and linked model evaluations were not validated. Illustrative thesis/genealogy scenarios are not executed case studies. Local/hosted model choice is not a privacy guarantee; README says reachable service pages are open as shipped until access control is enabled.","adoption_requirements":"README specifiesJava21+ (or bundled runtime), an OpenAI-compatible local/hosted model and a search backend; default fallback covers Wikipedia/Crossref/OpenAlex only. Inspect access controls/provider terms before restricted documents; no installation performed.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/researchzosho","sources":[{"label":"Wyrdsekai/researchzosho commit-pinned README","url":"https://raw.githubusercontent.com/Wyrdsekai/researchzosho/847200fc814076fe92cd46f3a3cb5fdeee2f3b09/README.md","kind":"code-repository","is_primary":1}]},{"slug":"reviewer-guideline-design-automated-peer-review","title":"Reviewer Guideline Design for Automated Peer Review","summary":"Experimental study of how reviewer-guideline design affects LLM-based automated peer review, reporting that official conference guidelines align most closely with human judgments.","resource_type":"workflow-process","disciplines":["computer-science"],"tasks":["automated-review"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans set objectives, supply data and check outputs.","source_checked_date":"2026-10-17","limitations":"Conference acceptance is recorded in the arXiv comment field; the venue version was not inspected. Experimental results are author-reported and not independently reproduced.","adoption_requirements":"Research paper; no tooling release inspected in checked sources.","updated_at":"2026-10-17T09:28:23Z","url":"https://research.wusong.net/cases/reviewer-guideline-design-automated-peer-review","sources":[{"label":"Evaluating the Impact of Reviewer Guideline Design on LLM-Based Automated Peer Review - arXiv 2607.22553 (ACL 2026 Findings)","url":"https://arxiv.org/abs/2607.22553","kind":"arxiv","is_primary":1}]},{"slug":"robin","title":"Robin: a multi-agent system for automating scientific discovery","summary":"A multi-agent system that integrates literature-search agents with data-analysis agents to automate hypothesis generation and data analysis for experimental biology. Applied to dry age-related macular degeneration, it proposed enhancing retinal pigment epithelium phagocytosis and identified and confirmed in vitro efficacy for two compounds, ripasudil and KL001.","resource_type":"use-case","disciplines":["biology","biomedicine"],"tasks":["data-analysis","drug-repurposing","hypothesis-generation","target-discovery"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":"L4","workflow_note":"Hypothesis generation and analysis iterate on experimental feedback. This is a bounded lab-in-the-loop workflow: humans execute wet-lab experiments, oversee safety and validate findings, not a fully autonomous laboratory.","source_checked_date":"2026-10-15","limitations":"Reported efficacy is in vitro only; no clinical trial or in vivo result is claimed. The workflow remains semi-autonomous: experiments are run by humans in the loop. All analyses and figures were produced by the system, so independent verification of the analysis pipeline is pending. Not independently reproduced.","adoption_requirements":"A working molecular-biology wet lab for validation; access to the system's code and foundation models; researchers willing to act as the experimental loop and validate system-proposed experiments.","updated_at":"2026-10-15T17:25:09Z","url":"https://research.wusong.net/cases/robin","sources":[{"label":"A multi-agent system for automating scientific discovery, Nature (2026)","url":"https://doi.org/10.1038/s41586-026-10652-y","kind":"doi","is_primary":1},{"label":"Code repository (Robin)","url":"https://github.com/Future-House/robin","kind":"code-repository","is_primary":0}]},{"slug":"robotreviewer","title":"RobotReviewer: assisting risk-of-bias assessment in clinical trial reports","summary":"A machine-learning system assesses risk of bias and extracts supporting text from clinical trial PDFs to assist evidence synthesis.","resource_type":"use-case","disciplines":["biomedicine","medicine"],"tasks":["knowledge-synthesis","literature-research"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":"L1","workflow_note":"Bounded task assistance: the model predicts risk-of-bias domains and selects supporting sentences from trial reports. Human systematic reviewers check the evidence and make final assessments; it does not conduct an entire review.","source_checked_date":"2026-10-15","limitations":"Model judgments were less accurate than published-review judgments. The evaluation concerns clinical trial bias assessment, not general research correctness; expert review remains necessary.","adoption_requirements":"Clinical trial PDFs, bias-assessment expertise, access to an appropriate implementation and human checks on predictions and evidence passages.","updated_at":"2026-10-15T17:25:09Z","url":"https://research.wusong.net/cases/robotreviewer","sources":[{"label":"Marshall, Kuiper & Wallace, RobotReviewer: evaluation of a system for automatically assessing bias in clinical trials, JAMIA (2016; online 2015)","url":"https://pmc.ncbi.nlm.nih.gov/articles/PMC4713900/","kind":"doi","is_primary":1}]},{"slug":"rs-ml-fpga-survey-data-workflow","title":"RS-ML-FPGA: survey data and analysis workflow for Earth-observation ML","summary":"A survey-linked collection of Earth-observation ML/FPGA studies, annotated research data, notebooks and table/figure generation code. The arXiv abstract reports 68 experiments and two taxonomies, while the project README documents the reusable analysis pipeline.","resource_type":"benchmark-dataset","disciplines":["computer-science","earth-environmental"],"tasks":["data-analysis","knowledge-synthesis","research-coding"],"evidence_level":"mixed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Workflow level intentionally unspecified for a reusable component or conceptual reference; researcher oversight and task-specific validation remain necessary.","source_checked_date":"2026-10-17","limitations":"Paper abstract plus the linked repository README reviewed, not the paper body, notebooks, underlying data or executable environment. The README states the original Zotero collection requires an API key. No code execution, corpus completeness, comparative FPGA performance or AI coding benefit was verified. The dataset link includes terminal punctuation in the source and was not followed.","adoption_requirements":"The README gives a Python/Conda environment setup and requirements-file installation. Reusing the original Zotero-connected pipeline requires collection identifiers, API access and the expected tag structure; inspect permissions and avoid committing private keys. Instructions were not executed.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/rs-ml-fpga-survey-data-workflow","sources":[{"label":"FPGA-Enabled Machine Learning Applications in Earth Observation: A Systematic Review","url":"https://arxiv.org/abs/2506.03938","kind":"arxiv","is_primary":1},{"label":"Primary repository README","url":"https://raw.githubusercontent.com/CedricLeon/Survey_RS-ML-FPGA/refs/heads/main/README.md","kind":"code-repository","is_primary":1}]},{"slug":"scienceagentbench","title":"ScienceAgentBench: task-level evaluation for scientific agents","summary":"A benchmark of publication-derived scientific tasks that evaluates generated programs rather than assuming end-to-end autonomous discovery.","resource_type":"benchmark-dataset","disciplines":["biomedicine","chemistry","earth-environmental","multidisciplinary","psychology"],"tasks":["agent-evaluation","data-analysis","research-coding"],"evidence_level":"project-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level entry: no universal autonomy level is assigned. The researcher defines the task, controls access and validates outputs; classify an actual instantiated workflow separately.","source_checked_date":"2026-10-15","limitations":"Benchmark performance cannot establish autonomous scientific discovery, validity in a new laboratory or general research productivity. No benchmark, repository code or generated program was executed by this directory.","adoption_requirements":"Benchmark datasets and harness, model/agent access, budget limits and an isolated execution environment. Dataset and software licenses require separate review.","updated_at":"2026-10-15T17:25:09Z","url":"https://research.wusong.net/cases/scienceagentbench","sources":[{"label":"Reviewed primary source: ScienceAgentBench: task-level evaluation for scientific agents","url":"https://osu-nlp-group.github.io/ScienceAgentBench/","kind":"project-documentation","is_primary":1},{"label":"Implementation linked by reviewed project documentation (not executed)","url":"https://github.com/OSU-NLP-Group/ScienceAgentBench","kind":"code-repository","is_primary":0}]},{"slug":"scientific-agent-skills","title":"Scientific Agent Skills: reusable research procedures, including literature review","summary":"A repository of reusable scientific task procedures for compatible AI agents. Review here covers the collection README and the literature-review SKILL.md, not an audit of every skill or an empirical test of research gains.","resource_type":"agent-skill","disciplines":["multidisciplinary"],"tasks":["knowledge-synthesis","literature-research","research-coding","workflow-orchestration"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"A library of instructions and optional helpers is not an autonomous workflow by itself. Level depends on the host agent, selected skill, granted tools and human checkpoints. Researchers must inspect skills before installation and approve data access, execution and final claims.","source_checked_date":"2026-10-15","limitations":"Maintainer documentation, not an independently tested claim of greater accuracy or faster research. Skill instructions can cause code execution, network requests and file changes in their host. The entire collection, dependencies and scripts were not audited; no downloaded skill was installed or executed.","adoption_requirements":"An Agent Skills-compatible host, narrowly scoped permissions, researcher oversight and task-specific dependencies. The reviewed literature-review skill declares Python 3.10+ with requests; optional PDF export requires Pandoc/XeLaTeX and optional discovery/generation services need their own access.","updated_at":"2026-10-15T17:25:09Z","url":"https://research.wusong.net/cases/scientific-agent-skills","sources":[{"label":"Scientific Agent Skills repository and README","url":"https://github.com/K-Dense-AI/scientific-agent-skills","kind":"code-repository","is_primary":1},{"label":"Reviewed literature-review SKILL.md (read as data; not installed)","url":"https://github.com/K-Dense-AI/scientific-agent-skills/blob/main/skills/literature-review/SKILL.md","kind":"project-documentation","is_primary":1},{"label":"Scientific Agent Skills LICENSE.md","url":"https://github.com/K-Dense-AI/scientific-agent-skills/blob/main/LICENSE.md","kind":"project-documentation","is_primary":1}]},{"slug":"scientific-generative-agent","title":"Scientific Generative Agent","summary":"Scientific Generative Agent alternates language-model proposals with simulation and continuous optimization. The paper evaluates four constitutive-law and four molecular-property tasks over five seeds, using task-specific losses. Its feedback loop is documented, but simulator and learned-property-model assumptions bound the conclusions. Public code is available; proposed molecules are not experimentally validated here.","resource_type":"framework","disciplines":["physics"],"tasks":["hypothesis-generation","optimization"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":"L3","workflow_note":"L3 applies only to the specified constitutive-law and molecular-property benchmark searches. Researchers define targets, valid bootstrapping designs and feedback summaries; the originating study compares task losses and analyses proposed solutions. Language-model proposals interact with differentiable simulation/parameter fitting; molecular properties use a learned surrogate. Eight tasks and five seeds define the reported evaluation scope. No experimental molecular validation or universal framework-level autonomy is established.","source_checked_date":"2026-10-19","limitations":"Simulator assumptions and a learned quantum-property surrogate constrain conclusions. Valid bootstrapping examples and human-designed feedback are supplied; no wet-lab or independent replication established.","adoption_requirements":"Untested suggestion: inspect the pinned sources and validate data, evaluator assumptions, implementation availability and researcher controls before planning a bounded trial. No setup or trial was executed.","updated_at":"2026-10-20T08:01:56Z","url":"https://research.wusong.net/cases/scientific-generative-agent","sources":[{"label":"Primary manuscript — 2405.09783v1","url":"https://arxiv.org/html/2405.09783v1","kind":"arxiv","is_primary":1},{"label":"Primary project repository or artifact (version limits apply)","url":"https://github.com/PingchuanMa/SGA/blob/a11a64e2abf26dd297bac88e8be7fa1fbffb4c31/README.md","kind":"code-repository","is_primary":1}]},{"slug":"sciscigpt","title":"SciSciGPT","summary":"SciSciGPT is an open-source prototype AI collaborator for science-of-science research described through empirical and analytical case studies.","resource_type":"framework","disciplines":["multidisciplinary","social-science"],"tasks":["data-analysis","literature-research","research-coding","workflow-orchestration"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Prototype case studies do not establish general research validity or independent replication. Source extract contains some truncated passages; no quantitative gains, universal autonomy level or reproduced results are claimed.","adoption_requirements":"Suitable research data, a configured model/tool environment and researchers able to inspect generated analyses and code.","updated_at":"2026-10-16T14:58:30Z","url":"https://research.wusong.net/cases/sciscigpt","sources":[{"label":"SciSciGPT: advancing human–AI collaboration in the science of science","url":"https://www.nature.com/articles/s43588-025-00906-6","kind":"doi","is_primary":1}]},{"slug":"scisen-scientific-writing-support","title":"SciSen: sentence-level scientific writing support","summary":"A sentence-level scientific-writing research workflow combining scientificness scoring, paper-section classification and paraphrasing; the paper links the SciSen companion repository.","resource_type":"workflow-process","disciplines":["computer-science","multidisciplinary"],"tasks":["scientific-writing"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Resource-level automation classification unspecified. This is a research/reference listing, not an exercised workflow; researchers select objectives and check interpretation and outputs.","source_checked_date":"2026-10-17","limitations":"Abstract-only substantive review and publisher publication headers, plus parent-proceedings metadata. Scientificness is a modelled writing-style signal, not scientific truth or research quality. Reported experiments were not independently reproduced; data splits, detailed metrics, bias results and hyperparameters were not inspected. The companion repository is linked by the primary abstract but was not installed or substantively reviewed. Direct Springer requests returned a JavaScript challenge; primary text was recovered using web_extract, which exposed no raw origin.","adoption_requirements":"Editorial, untested use: consult the paper and its explicitly linked SciSen repository before planning a pilot. Runtime dependencies, model weights, compute needs and data rights remain unverified. Researchers must check meaning preservation and confidential text handling.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/scisen-scientific-writing-support","sources":[{"label":"Fine-Tuning Language Models for Scientific Writing Support — arXiv","url":"https://arxiv.org/abs/2306.10974","kind":"arxiv","is_primary":1},{"label":"Springer — CD-MAKE 2023 chapter abstract and publication header","url":"https://link.springer.com/chapter/10.1007/978-3-031-40837-3_18","kind":"doi","is_primary":1}]},{"slug":"scrapesensellm","title":"ScrapeSenseLLM","summary":"Research tool documenting Reddit-data collection, local-LLM thematic coding using the Push-Pull-Mooring framework, and a dashboard for interpreting nootropics discourse.","resource_type":"tool","disciplines":["business","social-science"],"tasks":["data-analysis","data-extraction"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Workflow level intentionally unspecified for this reusable resource. Researchers select inputs, validate outputs and retain responsibility for interpretation.","source_checked_date":"2026-10-17","limitations":"README-only review; thesis, code and reliability calculations were not independently inspected or reproduced. README reports slight inter-rater agreement (kappa 0.115, 0.151 and 0.047); interpretive community archetypes must not be presented as reliable individual-level measurements. Several documentation links/parameters are blank and the balance formula has encoding corruption. No Reddit data is distributed according to README; privacy, consent/ethics and Reddit data terms require separate review.","adoption_requirements":"README demo specifies Python 3.11+, package installation and a synthetic-data generator followed by Streamlit. Real-data work additionally requires configured local-model tooling and authorized Reddit access; incomplete setup links need resolution.","updated_at":"2026-10-17T14:17:23Z","url":"https://research.wusong.net/cases/scrapesensellm","sources":[{"label":"Hotpopize/ScrapeSenseLLM primary repository documentation","url":"https://github.com/Hotpopize/ScrapeSenseLLM","kind":"code-repository","is_primary":1}]},{"slug":"scrapetx","title":"scrapeTx","summary":"A documented literature workspace connects topic and source selection, review, optional AI synthesis/extraction, editing and citation-linked document or slide export.","resource_type":"tool","disciplines":["multidisciplinary"],"tasks":["knowledge-synthesis","literature-research","scientific-writing"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Scoped documented L2: researchers choose topics/templates and sources, review literature, explicitly request bounded AI generation/extraction, then edit/export. This does not classify every gallery workflow or establish autonomous planning, docking validity or clinical effectiveness.","source_checked_date":"2026-10-19","limitations":"Commit-pinned README documentation only; repository code, linked guides, figures, dependencies and claimed tests were not executed/audited. No independent scientific or engineering result verified. Original queued repository Zntx-space now brands its README scrapeTx. Review is limited to the documented literature workflow, not endorsement of pharmaceutical/clinical/docking claims, security isolation, every gallery workflow or provider coverage. Optional AI uses configured providers; public search availability does not establish unrestricted full-text rights. README claims tests/local verification were not repeated on this Mac.","adoption_requirements":"README specifiesNode24 or22.13+, local SQLite fallback, optionalSupabase and user-configured AI/search integrations. Credentials/permissions and source rights are user responsibilities; no installation/provider connection performed.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/scrapetx","sources":[{"label":"ashwinsalunke/Zntx-space commit-pinned README","url":"https://raw.githubusercontent.com/ashwinsalunke/Zntx-space/7b11c4fc5d7a6a011c1ecc8bbf1323356463e3dd/README.md","kind":"code-repository","is_primary":1}]},{"slug":"scvi-tools","title":"scvi-tools","summary":"Probabilistic modeling toolkit for single-cell, multimodal and spatial omics built on PyTorch and AnnData.[13]","resource_type":"framework","disciplines":["biomedicine"],"tasks":["data-analysis"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"Models and preprocessing must fit the study design; not a generic claim of biological discovery. Documentation review only - no software execution or independent performance validation.","adoption_requirements":"Python; properly prepared omics data; method-specific assumptions and QC.","updated_at":"2026-10-17T06:33:49Z","url":"https://research.wusong.net/cases/scvi-tools","sources":[{"label":"scvi-tools - repository (scverse/scvi-tools)","url":"https://github.com/scverse/scvi-tools","kind":"code-repository","is_primary":1}]},{"slug":"sdl-demo-minimal-self-driving-laboratory","title":"SDL-Demo: minimal self-driving laboratory light-mixing demonstration","summary":"A minimal self-driving laboratory demonstration for teaching and prototyping closed-loop optimization: RGB light intensities are adjusted using measurements from a spectrophotometer to match a desired spectrum.","resource_type":"tutorial-template","disciplines":["chemistry","education","materials-science"],"tasks":["experiment-design","optimization"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level workflow is unspecified. Researchers set objectives, validate assumptions and outputs, and supervise any adoption.","source_checked_date":"2026-10-17","limitations":"Paper-excerpt review; tables, figures and linked build artifacts were not inspected or reproduced. Default setup has no robotic movement. Its simple, largely linear response surface is not a representative benchmark for complex materials discovery; historical cost/setup targets are not verified current purchasing estimates.","adoption_requirements":"Compatible LED, spectrophotometer, microcontroller and software setup; wiring, component tests and researcher-defined objectives. Review remote-access/cloud settings and hardware safety before use.","updated_at":"2026-10-17T14:30:44Z","url":"https://research.wusong.net/cases/sdl-demo-minimal-self-driving-laboratory","sources":[{"label":"What is a minimal working example for a self-driving laboratory?","url":"https://doi.org/10.1016/j.matt.2022.11.007","kind":"doi","is_primary":1}]},{"slug":"sine-gordon-localization","title":"Sine-Gordon energy-localization prediction","summary":"A documented computational-physics workflow connects simulated pendulum-ring data, physics-informed CNN design, model training and ablation/robustness evaluation to predict where energy localizes. Results are author-reported simulations.","resource_type":"use-case","disciplines":["physics"],"tasks":["data-analysis","optimization"],"evidence_level":"repository-documented","outcome_verification":"author-reported","workflow_level":"L2","workflow_note":"Scoped L2 for the repository-described simulation-pool, CNN training/validation and ablation process. Researchers define the problem, datasets and comparisons; author-reported simulation results are not independently reproduced or physical deployment evidence.","source_checked_date":"2026-10-19","limitations":"Repository-documented, author-reported simulation study; neither software nor outcomes independently reproduced. v2 comparisons use a rebuilt v1 (60.4% within ±1), not the unavailable original pipeline reported at94.4%; do not claim v2 surpasses the published original. Camera/noise scenarios are simulations, with a bench-top test proposed. Idealised independent sensor noise, linear uniform damping, limited impurity tests, different seed/screening budgets and label definitions limit transfer. Publisher DOI was inaccessible through the tool; no full publisher-paper or peer-review audit. Saved REPORT coverage is selective.","adoption_requirements":"Source documents Python3.12 and PyTorch training on Apple MPS, NVIDIA CUDA or CPU. Generated pools/checkpoints are not tracked and take severalGB; estimate resources and inspect dependencies before attempting reproduction. Installation and runtime compatibility were not tested.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/sine-gordon-localization","sources":[{"label":"Commit-pinned repository documentation","url":"https://raw.githubusercontent.com/adrian0729/sineGordonLocalization/eecd21cdaba4a1e31d8b7d45b63dc4520cb7e282/README.md","kind":"code-repository","is_primary":1},{"label":"REPORT Methods, lines224-227","url":"https://raw.githubusercontent.com/adrian0729/sineGordonLocalization/eecd21cdaba4a1e31d8b7d45b63dc4520cb7e282/REPORT.md","kind":"code-repository","is_primary":1},{"label":"Scripts phases introductory paragraph","url":"https://raw.githubusercontent.com/adrian0729/sineGordonLocalization/eecd21cdaba4a1e31d8b7d45b63dc4520cb7e282/scripts/README.md","kind":"code-repository","is_primary":1}]},{"slug":"skillspector-snapshot-security-analysis-reference","title":"SkillSpector: documented skill-security analysis and triage","summary":"SkillSpector documents static and optional LLM-assisted analysis of agent-skill contents, followed by risk reports and user-defined triage.","resource_type":"tool","disciplines":["computer-science"],"tasks":["agent-evaluation","data-analysis"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Intended documented L2 for one SkillSpector reference: users select inputs and mode; static and optional semantic stages produce reports; users accept known findings and set gate policy. Capabilities are bound to the original GitLab README snapshot. Related copies are not separate products, and current upstream differences are not merged into that snapshot.","source_checked_date":"2026-10-19","limitations":"Queued GitLab README/metadata and selected current upstream pipeline/trust/limitation passages only. Moving HEADs are unpinned and non-atomic. The queued snapshot says semantic analysis filters false positives and recognised signature bundles are excluded; upstream instead describes retaining deterministic findings and analysing DSSE contents. No official GitLab affiliation, distinct fork contribution or current implementation certification is inferred. Earlier related-copy matches do not prove present equivalence. Headline dataset vulnerability rates and precision estimates were not validated. LLM mode sends file contents to its configured provider; static-only mode can still send dependency coordinates to OSV. No sandbox, signature validity, complete coverage, safety or runtime assurance is established. No scanner, model or provider was run.","adoption_requirements":"Untested suggestion: select and pin a source revision, confirm finding-retention/signature/completeness behaviour, check egress and data rights, and validate reports against known findings before a research trial.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/skillspector-snapshot-security-analysis-reference","sources":[{"label":"Original queued GitLab README snapshot","url":"https://gitlab.com/motherskitchenblr2/SkillSpector/-/raw/HEAD/README.md","kind":"code-repository","is_primary":1},{"label":"Upstream README snapshot: differing pipeline and trust notes","url":"https://raw.githubusercontent.com/NVIDIA/skillspector/HEAD/README.md","kind":"code-repository","is_primary":1}]},{"slug":"soss-loom-active-creative-medium","title":"SOSS and Loom: generative AI as an active creative medium","summary":"A creative-process framework, SOSS (Shape, Observe, Stir and Select), treats generative AI as an active medium rather than only an advisor. The authors introduce Loom, a creative-writing probe in which users orchestrate simulated narrative agents.","resource_type":"workflow-process","disciplines":["arts","computer-science"],"tasks":["workflow-orchestration"],"evidence_level":"mixed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level workflow level intentionally unspecified. Conceptual references and reusable process patterns do not themselves establish an autonomous workflow; researchers retain task-specific interpretation and oversight.","source_checked_date":"2026-10-17","limitations":"Abstract-only substantive review with official workshop acceptance-list metadata. The full Loom design, implementation, access terms, prompts, experiments and usability were not read or tested. A CHI 2026 Tools for Thought non-archival workshop contribution is confirmed; the site's acceptance list does not itself establish peer-review status. Mixed denotes arXiv abstract and official workshop documentation, not confirmed archival peer review. Claimed creative quality and convergence tendencies are the authors' arguments, not reproduced outcomes.","adoption_requirements":"Editorial, untested: obtain the complete description and establish whether the probe is accessible before attempting implementation. Use material with appropriate rights and protect private writing; neither a software distribution nor a working installation was verified.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/soss-loom-active-creative-medium","sources":[{"label":"Material for Thought — primary arXiv v1 abstract","url":"https://arxiv.org/abs/2605.19832","kind":"arxiv","is_primary":1},{"label":"CHI 2026 Tools for Thought — official accepted-submission list","url":"https://ai-tools-for-thought.github.io/workshop/","kind":"workshop-page","is_primary":1}]},{"slug":"spire","title":"SPIRE","summary":"SPIRE coordinates seven scholarly agents that retrieve and compare primary passages, bind claims to evidence, and request further retrieval through bounded reflection. Its author study covers classical Chinese and Latin scholarship; citation recovery and essay ratings do not establish interpretive truth. The linked repository returned 404, so code availability and experiment-version parity remain unverified.","resource_type":"framework","disciplines":["humanities"],"tasks":["archival-research","knowledge-synthesis"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":"L3","workflow_note":"L3 applies only to the corpus- and question-bounded evidence-grounded scholarship workflow. A user supplies a classical-text corpus, schema and question. Scholars retain interpretive judgement; blind human ratings are within the author study, not an installed approval mechanism. Seven roles retrieve, compare and bind evidence; reflection may request further retrieval. Reflection stops when its review passes or at two rounds. Citation recovery and study ratings do not certify historical interpretation or replace expert source judgement.","source_checked_date":"2026-10-19","limitations":"Chinese/Latin textual corpora only; citation recovery is not unique interpretive truth. Translation, scope and locus failures remain. Human ratings are within the author study, not independent replication. Dataset redistribution is restricted/derived; raw texts are not promised.","adoption_requirements":"Untested suggestion: inspect the pinned sources and validate data, evaluator assumptions, implementation availability and researcher controls before planning a bounded trial. No setup or trial was executed.","updated_at":"2026-10-20T08:01:56Z","url":"https://research.wusong.net/cases/spire","sources":[{"label":"Primary manuscript — arXiv2605.30947v4;31August2026","url":"https://arxiv.org/html/2605.30947v4","kind":"arxiv","is_primary":1}]},{"slug":"spontaneous-emission-self-driving-lab","title":"Self-driving laboratory for steering spontaneous emission","summary":"An ultrafast nanophotonics platform combines generative models, active learning and equation learning to study steering of spontaneous emission from reconfigurable semiconductor metasurfaces.","resource_type":"framework","disciplines":["materials-science","physics"],"tasks":["data-analysis","experiment-design","experiment-execution","optimization"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Platform-level automation is author-described; workflow level left unspecified because scope and researcher responsibilities depend on the implementation. Human safety and research oversight remain required.","source_checked_date":"2026-10-17","limitations":"Abstract-only canonical review of the current arXiv record; an auxiliary paper extraction was only partially read. Numerical performance differs between that extraction and the current abstract, so no numerical performance claim is retained. No hardware, code, equations or experimental outcomes independently reproduced; generalization beyond the studied metasurfaces is unverified.","adoption_requirements":"Editorial, untested guidance: assess access to reconfigurable metasurfaces, ultrafast optical measurement and ML expertise; review apparatus and calibration requirements before attempting reproduction, with trained laser-safety oversight.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/spontaneous-emission-self-driving-lab","sources":[{"label":"Self-driving lab discovers principles for steering spontaneous emission","url":"https://arxiv.org/abs/2407.16083","kind":"arxiv","is_primary":1}]},{"slug":"ssh-citation-index-modules","title":"SSH Citation Index Modules","summary":"AI modules for extracting, parsing and disambiguating bibliographic references in social-science and humanities publications.","resource_type":"framework","disciplines":["humanities","social-science"],"tasks":["data-extraction","data-management","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"A defined extraction/parsing/linking workflow, not autonomous scholarship. Researchers must inspect extraction completeness, bibliographic matches and ambiguous links.","source_checked_date":"2026-10-17","limitations":"README explicitly discourages production use while under development. Citation-intent classification and entity extraction are TODO, not delivered capabilities. Benchmark directories and mocked linking tests do not establish accuracy; no benchmark or live workflow was reproduced. Documented requirements.txt install command should be checked against checkout because the retrieved root file listing does not show that file.","adoption_requirements":"Python 3.10+ for development; documented API deployment requires Docker services, Redis/workers, an OpenAI-compatible vLLM endpoint and external MinerU service when used.","updated_at":"2026-10-17T11:14:58Z","url":"https://research.wusong.net/cases/ssh-citation-index-modules","sources":[{"label":"SSH Citation Index Modules — primary repository","url":"https://github.com/odoma-ch/ssh-citation-index","kind":"code-repository","is_primary":1}]},{"slug":"storm","title":"STORM: a framework for source-grounded pre-writing and knowledge curation","summary":"A modular LLM framework gathers information through retrieval and multi-perspective question asking, then produces outlines and cited drafts. It is a research pre-writing starting point, not a validated end-to-end academic writing service.","resource_type":"framework","disciplines":["computer-science","multidisciplinary"],"tasks":["knowledge-synthesis","literature-research","scientific-writing"],"evidence_level":"mixed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Reusable framework, so no universal L1–L4 label is assigned. Its documented STORM implementation performs multi-step retrieval, question asking, outlining and writing; humans supply the topic, configure sources/models and check citations and claims. An instantiated research workflow needs its own classification.","source_checked_date":"2026-10-15","limitations":"The authors explicitly say outputs are not publication-ready. Source bias, unsupported associations and incomplete retrieval remain concerns. A peer-reviewed evaluation of Wikipedia-like articles must not be relabeled as proven academic benefit.","adoption_requirements":"Python development environment, a configured language model and retrieval source, API/corpus access where needed, budget controls, and human verification of sources and generated prose.","updated_at":"2026-10-15T17:25:09Z","url":"https://research.wusong.net/cases/storm","sources":[{"label":"STORM repository and README","url":"https://github.com/stanford-oval/storm","kind":"code-repository","is_primary":1},{"label":"Shao et al., Assisting in Writing Wikipedia-like Articles From Scratch with Large Language Models, NAACL 2024","url":"https://aclanthology.org/2024.naacl-long.347/","kind":"doi","is_primary":1},{"label":"STORM MIT LICENSE","url":"https://github.com/stanford-oval/storm/blob/main/LICENSE","kind":"project-documentation","is_primary":1}]},{"slug":"stride-concept-drift-explanation","title":"STRIDE: stream insight and drift explanation","summary":"A documented Python toolkit and interactive dashboard connects streaming data, model-health/drift signals and model-explanation views for exploratory concept-drift research.","resource_type":"tool","disciplines":["computer-science"],"tasks":["data-analysis"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Intended L2 for researcher-configured streams/windows/models, drift analysis and explanation/dashboard comparison. Statistical detectors and xAI do not establish adaptive tool-planning agency, autonomous scientific iteration or causal explanation. No implementation or research outcome was evaluated.","source_checked_date":"2026-10-19","limitations":"Full README read; no code, diagrams, hosted dashboard, linked paper/PDF, installation or model execution. README claims ECML PKDD2026 DemoTrack presentation and gives DOI; this source check does not independently verify peer review or research outcomes. No empirical benefit or accuracy comparison accepted. Statistical significance and explanation tools do not establish causal validity; analyst validation remains necessary. Not the unrelated STRIDE security acronym. AI editorial source review only; no installation, source code execution, provider/model calls or outcome reproduction.","adoption_requirements":"README specifies Python3.10–3.12 and stride-xai with optional dashboard/xAI extras; installation and live dependency compatibility are untested. Check data rights and analysis assumptions before a separate trial.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/stride-concept-drift-explanation","sources":[{"label":"KubaCzech/STRIDE primary-readme","url":"https://raw.githubusercontent.com/KubaCzech/STRIDE/794ff769ddfa187ab08a7a3e0b213345fe9ed6ae/README.md","kind":"code-repository","is_primary":1}]},{"slug":"structured-prompt-engineering-inductive-thematic-analysis","title":"Structured prompt engineering for LLM-assisted inductive thematic analysis","summary":"Khalid and Witmer review LLM-assisted inductive thematic analysis and propose a structured four-step prompt-engineering process, with discussion of advanced prompting techniques and research directions. This listing describes the proposal, not a reproduced recipe or validated performance gain.","resource_type":"workflow-process","disciplines":["social-science"],"tasks":["data-analysis"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Researcher-supervised process; automation level intentionally unspecified. No autonomous research or clinical deployment asserted.","source_checked_date":"2026-10-17","limitations":"Abstract-only review of the author-institution Illinois Experts record; publisher DOI extraction returned HTTP403. The four steps, exact prompts, models, datasets and any quantitative evaluation were not available in the reviewed abstract. Benefits are methodological intentions, not verified outcomes.","adoption_requirements":"Qualitative-methods expertise, access to the detailed method, an institutionally permitted LLM environment and consent/privacy safeguards for interview data. No code, model subscription or open licence established by this abstract.","updated_at":"2026-10-17T18:33:50Z","url":"https://research.wusong.net/cases/structured-prompt-engineering-inductive-thematic-analysis","sources":[{"label":"Prompt Engineering for Large Language Model-Assisted Inductive Thematic Analysis","url":"https://experts.illinois.edu/en/publications/prompt-engineering-for-large-language-model-assisted-inductive-th/","kind":"institutional-report","is_primary":1}]},{"slug":"swif2t-focused-scientific-feedback","title":"SWIF²T: focused feedback for scientific writing","summary":"SWIF²T is an LLM-based focused-feedback framework intended to identify weaknesses in scientific manuscripts and suggest revisions through planner, investigator, reviewer and controller components.","resource_type":"framework","disciplines":["computer-science","multidisciplinary"],"tasks":["automated-review","scientific-writing"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Resource-level automation classification unspecified. This is a research/reference listing, not an exercised workflow; researchers select objectives and check interpretation and outputs.","source_checked_date":"2026-10-17","limitations":"Abstract-only substantive review plus official ACL Anthology publication metadata. Dataset size is not the number of human evaluators or a count of tested papers. Human evaluation belongs to the authors’ study, not this AI editorial review. Model choices, prompts, scoring protocol, statistical significance and failure rates were not inspected. No generated feedback, software execution, privacy/security audit or independent outcome reproduction was performed. Reported helpfulness does not establish manuscript correctness or justify replacing expert review.","adoption_requirements":"Editorial, untested use: read the paper and obtain implementation and evaluation details before any pilot. Runtime, model-provider costs and repository distribution were not established in this abstract review. Do not expose confidential or unpublished work without an appropriate data-handling basis.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/swif2t-focused-scientific-feedback","sources":[{"label":"SWIF²T — arXiv abstract","url":"https://arxiv.org/abs/2405.20477","kind":"arxiv","is_primary":1},{"label":"ACL Anthology — official Findings ACL 2024 publication","url":"https://aclanthology.org/2024.findings-acl.580/","kind":"doi","is_primary":1}]},{"slug":"swiss-education-llm-adoption-criteria","title":"Educational LLM adaptation in German and Swiss German: contextual and governance criteria","summary":"Author-hosted applied-research report offering contextual-material, language-fit and responsible-deployment criteria for educational LLM adoption in German/Swiss German settings.","resource_type":"workflow-process","disciplines":["education","linguistics"],"tasks":["knowledge-synthesis"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation left unspecified. Researchers/educators choose materials, govern model access and verify interpretations and outputs.","source_checked_date":"2026-10-17","limitations":"Author-hosted README plus selected report pages 1, 4–6 and 9 reviewed, not a full-text claim audit. This is a self-described master’s applied-research report, not verified peer-reviewed publication. The admitted content is limited to educational adoption criteria, literature-review/interview method and governance questions. Historical/technical attributions, vendor assurances, market claims and secondary performance figures are not endorsed; the technical-history narrative should be checked against originals before reuse. Fine-tuning is future work within the report; the separate lyrics-generator follow-up was not reviewed. No learning gain or model effectiveness evaluated. The report’s technical-history section includes unreliable attributions; do not use this listing as technical instruction.","adoption_requirements":"No software installation is needed to read the report. Actual adoption requires permitted educational materials, relevant language expertise and local privacy/integrity review; these are editorial implementation considerations, not an executed recipe.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/swiss-education-llm-adoption-criteria","sources":[{"label":"sreterannamaria/llm-swiss-education-paper — primary README","url":"https://github.com/sreterannamaria/llm-swiss-education-paper","kind":"code-repository","is_primary":1},{"label":"Fischer, Schmid and Sréter — report excerpts, pages 1, 4–6 and 9","url":"https://raw.githubusercontent.com/sreterannamaria/llm-swiss-education-paper/main/AFE1_LLM_Paper.pdf","kind":"institutional-report","is_primary":1}]},{"slug":"symdis","title":"SyMDis","summary":"Sample-efficient symbolic-learning optimisation method that discovers near-optimal materials from large databases and learns interpretable rules.[13]","resource_type":"framework","disciplines":["computer-science","materials-science"],"tasks":["materials-discovery","optimization"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans set objectives and validate results.","source_checked_date":"2026-10-16","limitations":"Preprint; no code availability verified; not independently reproduced.","adoption_requirements":"A database of candidate materials with a scoring or simulation function; no packaged implementation verified.","updated_at":"2026-10-17T08:00:32Z","url":"https://research.wusong.net/cases/symdis","sources":[{"label":"Symbolic Learning for Material Discovery - arXiv:2312.11487","url":"https://arxiv.org/abs/2312.11487","kind":"arxiv","is_primary":1}]},{"slug":"synthetic-biology-sdl-digital-twin-perspective","title":"Synthetic-biology SDLs: prospective digital-twin research framework","summary":"A perspective describes a proposed network of biological SDLs, human researchers and evolving digital twins.","resource_type":"framework","disciplines":["biology","engineering"],"tasks":["experiment-design","workflow-orchestration"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level assigned. This prospective reference connects experiments, evolving digital twins and human/laboratory knowledge exchange. An implemented system and specific human approval/validation protocol are not established; prospective autonomy terminology does not assign a catalogue level.","source_checked_date":"2026-10-19","limitations":"Selected PDF passages and official abstract only. Figures, references, cited experiments and implementations were not audited. The authors’ statement about unavailable technology is an archival prospective claim, not a current availability check. Cached extraction/raw-version parity remains unverified. No biological result or operational laboratory instruction is inherited.","adoption_requirements":"Untested suggestion: inspect the original proposal and separate prospective components from established methods before planning research.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/synthetic-biology-sdl-digital-twin-perspective","sources":[{"label":"Selected perspective passages","url":"https://arxiv.org/pdf/2210.09085","kind":"arxiv","is_primary":1}]},{"slug":"tabular-adversarial-research-taxonomy","title":"Tabular adversarial-ML research: strategy and evaluation taxonomy","summary":"A review categorises tabular adversarial-ML methods and how studies address practical evaluation requirements.","resource_type":"framework","disciplines":["computer-science"],"tasks":["experiment-design","literature-research"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level assigned. The review categorises tabular adversarial-ML method families and practical evaluation dimensions. Its human-authored categorisation does not establish one AI-assisted research workflow or validation protocol.","source_checked_date":"2026-10-19","limitations":"Selected HTML method, taxonomy, evaluation and discussion passages only. The complete paper, appendices, primary studies, code, datasets and numerical findings were not audited. Firstness, completeness and attack-count claims are not validated. Feasibility, semantics and plausibility are inconsistently defined, so source categories do not prove real-world validity. No attack procedure, security-system change or experiment was performed.","adoption_requirements":"Untested suggestion: use the taxonomy to locate primary studies and check definitions, constraints, benchmarks and dataset transfer before research reuse.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/tabular-adversarial-research-taxonomy","sources":[{"label":"Selected taxonomy and evaluation passages","url":"https://arxiv.org/html/2506.15506","kind":"arxiv","is_primary":1}]},{"slug":"tacit-qualitative-coding","title":"TACIT: theory-anchored qualitative coding","summary":"Streamlit application documented for LLM-assisted interview coding, researcher revision and an audit trail.","resource_type":"tool","disciplines":["multidisciplinary","social-science"],"tasks":["data-analysis","data-management"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation is unspecified; researchers choose sources, approve interpretations and remain responsible for results.","source_checked_date":"2026-10-17","limitations":"README-only review; no code, paper or benchmark reproduction. Synthetic reference codings are not empirical validation. Model omissions and unsupported interpretation require researcher review. Local transcript processing depends on a correctly configured local endpoint; literature retrieval and cloud providers can use the network.","adoption_requirements":"Python 3.9+ and dependencies; a supported local model server or Gemini access for model-assisted coding. Human-subject approval and appropriate endpoint/data handling are required.","updated_at":"2026-10-17T16:34:24Z","url":"https://research.wusong.net/cases/tacit-qualitative-coding","sources":[{"label":"NYCU-FUN-TECH-LAB/TACIT — primary README","url":"https://github.com/NYCU-FUN-TECH-LAB/TACIT","kind":"code-repository","is_primary":1}]},{"slug":"talkmoves","title":"TalkMoves: transformer-assisted classroom discourse analysis","summary":"A research application combining speech recognition and transformer models to analyze mathematics classroom talk moves.","resource_type":"tool","disciplines":["education"],"tasks":["data-analysis","transcription"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level entry: no universal autonomy level is assigned. The researcher defines the task, controls access and validates outputs; classify an actual instantiated workflow separately.","source_checked_date":"2026-10-15","limitations":"The retrieved ar5iv representation is short: abstract and introductory model description only. Full evaluation, classroom generalization, deployment availability and reuse license were not checked. No claims of improved student learning or independently replicated model performance are made.","adoption_requirements":"Classroom recordings/transcripts, participant consent, a suitable discourse coding scheme and expert checking. Current service or code availability is not verified.","updated_at":"2026-10-15T17:25:09Z","url":"https://research.wusong.net/cases/talkmoves","sources":[{"label":"Reviewed primary source: TalkMoves: transformer-assisted classroom discourse analysis","url":"https://ar5iv.labs.arxiv.org/html/2105.07949","kind":"arxiv","is_primary":1}]},{"slug":"tc-conet","title":"TC-CoNet","summary":"TC-CoNet provides PyTorch code and configuration for 3D medical-image segmentation research, with README instructions for preprocessing, training and inference.","resource_type":"tool","disciplines":["biomedicine"],"tasks":["image-analysis","research-coding"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Workflow level intentionally unspecified for this reusable resource. Researchers select inputs, validate outputs and retain responsibility for interpretation.","source_checked_date":"2026-10-17","limitations":"Primary README only; associated journal paper, training data and code were not evaluated. No segmentation accuracy or clinical utility is verified. Documented environment is old (Python 3.7, PyTorch 1.7.1, Windows 10); present compatibility is unknown. Dataset permissions and sensitive medical-data governance require separate review.","adoption_requirements":"README specifies Python 3.7, PyTorch 1.7.1 and Windows 10 and refers to nn-UNet for dataset configuration. Dataset access, appropriate medical-data permissions and a compatible environment are required.","updated_at":"2026-10-17T14:17:23Z","url":"https://research.wusong.net/cases/tc-conet","sources":[{"label":"YongChen-Exact/TC-CoNet primary repository documentation","url":"https://github.com/YongChen-Exact/TC-CoNet","kind":"code-repository","is_primary":1}]},{"slug":"tech-research-reading-digest-workflow","title":"Tech Research: AI-assisted reading and digest workflow","summary":"A personal repository documenting an AI-assisted technology-reading workflow focused on AI cognition and the technology–humanities intersection. It combines a subscription configuration, agent-assisted source scanning and a weekly digest scaffold, while leaving substantive reading annotations to its owner.","resource_type":"workflow-process","disciplines":["humanities","psychology"],"tasks":["knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No fixed workflow level. Agents assist discovery and scaffolding; the owner reads and supplies substantive notes. Any academic adaptation requires researchers to select sources, check claims and separate weak signals from evidence.","source_checked_date":"2026-10-17","limitations":"README-only review; scheduler activity, source configuration, generator code, digests and linked research notes were not verified. Sources include blogs, newsletters and social signals, not a controlled scholarly search. The two-source trigger is a discovery heuristic, not independent confirmation. No synthesis quality, coverage, research result or productivity outcome was evaluated. AI retrieval does not fill the owner’s substantive interpretation automatically.","adoption_requirements":"The documented local scaffold uses Python and a subscription YAML configuration; the agent retrieval schedule depends on the owner’s QoderWork environment and is not supplied as a verified portable service. Check source terms, copyright, privacy and model-provider exposure; retain source attribution and researcher review.","updated_at":"2026-10-17T21:43:09Z","url":"https://research.wusong.net/cases/tech-research-reading-digest-workflow","sources":[{"label":"Tech Research: AI-assisted reading and digest workflow — repository README","url":"https://github.com/yithmon/tech-research","kind":"code-repository","is_primary":1}]},{"slug":"the-rookery-storm-research","title":"The Rookery: Storm Research skill","summary":"The Rookery documents a portable Storm Research skill for source-backed briefings using independent perspectives, explicit disagreements, and source-quality audits. It is an independent workflow inspired by STORM, not a distribution of Stanford STORM software.","resource_type":"agent-skill","disciplines":["multidisciplinary"],"tasks":["knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No fixed automation level is assigned. The documented agent workflow proposes and synthesizes evidence; researchers remain responsible for framing, source rights, disciplinary judgment and final use.","source_checked_date":"2026-10-17","limitations":"README and Storm Research SKILL.md only; linked lens charter, analysis-methods and fidelity-check reference files were not reviewed. No installation, research run or behavioral evaluation was performed. The README says other skill packages await revalidation after an evaluation reset. Multi-agent separation depends on the host harness and does not establish independent scientific evidence. No accuracy or time-saving outcome is established.","adoption_requirements":"A compatible Agent Skills host and permitted source access are needed; the documented isolated-lens mode requires verifiably clean executor contexts. Audit package instructions, permissions and model-provider data handling before use; do not upload confidential research material without authorization.","updated_at":"2026-10-17T21:43:09Z","url":"https://research.wusong.net/cases/the-rookery-storm-research","sources":[{"label":"The Rookery: Storm Research skill — repository README","url":"https://github.com/jrgilbertson/the-rookery","kind":"code-repository","is_primary":1},{"label":"The Rookery: Storm Research skill — targeted primary documentation","url":"https://raw.githubusercontent.com/jrgilbertson/the-rookery/HEAD/skills/storm-research/SKILL.md","kind":"project-documentation","is_primary":1}]},{"slug":"theory-first-generative-ai-humanities","title":"A theory-first approach toward using generative AI for humanities research","summary":"Andrew Piper proposes theory-first guidance for generative-AI-assisted humanities research: explicitly define research problems and analytical constructs, distinguish AI as an object from AI as an instrument, and use theoretical commitments to guide evaluation and interpretation. The article argues for methodological rigor rather than demonstrating an independently verified performance gain.","resource_type":"workflow-process","disciplines":["humanities","social-science"],"tasks":["data-analysis","experiment-design"],"evidence_level":"project-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation unspecified; researchers define aims and verify outputs. No autonomous end-to-end outcome claimed.","source_checked_date":"2026-10-17","limitations":"Publisher abstract, plain-language summary, introduction and problem-formation excerpts reviewed; not a full-text adjudication despite the archived HTML containing further text. Examples and cited studies were not independently verified. No software execution or outcome reproduction.","adoption_requirements":"A humanities research question, explicit theoretical commitments, suitable texts or cultural materials and domain expertise. Any model use also requires appropriate data rights, privacy review and documented model/prompt choices.","updated_at":"2026-10-17T20:06:54Z","url":"https://research.wusong.net/cases/theory-first-generative-ai-humanities","sources":[{"label":"A theory-first approach toward using generative AI for humanities research","url":"https://doi.org/10.1017/chr.2026.10039","kind":"doi","is_primary":1}]},{"slug":"trace-archives","title":"TRACE archives","summary":"TRACE plans, searches, reviews and reranks historical archive sources. The author evaluation uses 1,752 questions over French documents from 1887; retrieval recall measures source discovery, not historical truth. Human checkpoints and expert-trust benefits remain future work. A public implementation is documented, but its match to the evaluated version and independent replication are unverified.","resource_type":"framework","disciplines":["humanities"],"tasks":["archival-research","literature-research"],"evidence_level":"peer-reviewed","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"No workflow level assigned: the archival source-discovery algorithm. The researcher query/corpus is the input; expert checkpoints and trust/verification-effort benefits are proposed future work. Planning, retrieval, review and reranking are documented, but a completed human review/override pathway for the particular scholarly use is not established. Specific missing evidence: source-grounded operator validation/override and responsibility for final historical-source selection in an instantiated workflow. Review later primary workflow documentation before assigning a level; keep the bounded retrieval-method entry without an invented level.","source_checked_date":"2026-10-19","limitations":"Retrieval, not historical-interpretation quality. Baseline trajectory discovery-order ranking is asymmetric; single-year corpus/constants limit transfer. Internal24researcher/sixinstitution availability is author-reported, not user-study evidence; trust/verification-effort benefits remain untested.","adoption_requirements":"Untested suggestion: inspect the pinned sources and validate data, evaluator assumptions, implementation availability and researcher controls before planning a bounded trial. No setup or trial was executed.","updated_at":"2026-10-20T08:01:56Z","url":"https://research.wusong.net/cases/trace-archives","sources":[{"label":"Primary manuscript — arXiv2609.19897v1;17September2026","url":"https://arxiv.org/html/2609.19897v1","kind":"arxiv","is_primary":1},{"label":"Primary project repository or artifact (version limits apply)","url":"https://github.com/Kepler1908/TRACE","kind":"code-repository","is_primary":1}]},{"slug":"traceableai","title":"TraceableAI","summary":"Open-source application that automates a PRISMA 2020 literature-review workflow so that every citation in the generated review can be traced back to a retrieved source.","resource_type":"tool","disciplines":["computer-science","multidisciplinary"],"tasks":["automated-review","literature-research"],"evidence_level":"repository-documented","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Automates the PRISMA 2020 sequence with traceable, checkable citations.","source_checked_date":"2026-10-17","limitations":"Author-described PRISMA workflow; not executed. Saved README excerpts do not establish complete setup requirements or independent effectiveness.","adoption_requirements":"Repository indicates .NET 10.0 (C#); consult current setup documentation for provider credentials and costs. Apache-2.0 code; bundled SCImago data has separate non-commercial attribution terms.","updated_at":"2026-10-17T09:28:23Z","url":"https://research.wusong.net/cases/traceableai","sources":[{"label":"au-btech-literature-review-agent - GitHub repository (lauPhilip)","url":"https://github.com/lauPhilip/au-btech-literature-review-agent","kind":"code-repository","is_primary":1}]},{"slug":"tracereview","title":"TraceReview: source-matched literature reading and audit trail","summary":"A documented literature application connects discovery and AI-assisted summaries with source-matched excerpts and an inspectable workflow audit trail.","resource_type":"tool","disciplines":["multidisciplinary"],"tasks":["knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Intended L2 for question-driven search, deduplication/lexical ranking, model summaries and excerpt verification with source/audit inspection. Wording matching does not verify scientific truth. No adaptive planning, implemented embedding ranking or resumable-run claim.","source_checked_date":"2026-10-19","limitations":"Full README and raw workflow/feature SVG source text read; no rendering/JS, app/code/model, Docker, provider or database executed. Wording matching/fallback does not establish truth or coverage. Source diagram says paper cards and a report, but README lists exportable Markdown/PDF reports as future; no export capability claimed. Active development, deterministic lexical ranking and restart non-resumption explicit. Licence file not audited and no licence prerequisite imposed. Moving default-branch source timestamps do not establish commit/atomic-version consistency. AI editorial review only; no source code, models or installation executed, no production/status/publication changes.","adoption_requirements":"README names Docker Compose, Groq model configuration and PostgreSQL credentials; inspect current dependencies/provider/data handling and source rights before any separate trial. No credentials, app or services touched.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/tracereview","sources":[{"label":"samavinasir/tracereview primary-readme","url":"https://raw.githubusercontent.com/samavinasir/tracereview/HEAD/README.md","kind":"code-repository","is_primary":1},{"label":"samavinasir/tracereview workflow-svg","url":"https://raw.githubusercontent.com/samavinasir/tracereview/HEAD/assets/workflow.svg","kind":"project-documentation","is_primary":1},{"label":"samavinasir/tracereview features-svg","url":"https://raw.githubusercontent.com/samavinasir/tracereview/HEAD/assets/features.svg","kind":"project-documentation","is_primary":1}]},{"slug":"transkribus","title":"Transkribus: AI-assisted transcription for archival research","summary":"A hosted platform documenting handwriting recognition, correction, extraction and publication of archival material.","resource_type":"tool","disciplines":["humanities"],"tasks":["archival-research","data-extraction"],"evidence_level":"vendor-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level entry: no universal autonomy level is assigned. The researcher defines the task, controls access and validates outputs; classify an actual instantiated workflow separately.","source_checked_date":"2026-10-15","limitations":"Vendor descriptions of speed, language coverage and reproducibility were not independently evaluated. Historical scripts, degraded documents and layout variation require manual checking. Current prices and service/privacy terms were not audited.","adoption_requirements":"Document images, rights and consent to process them, a suitable recognition model, human correction and a service account. Hosted cooperative/commercial service; no purchase made.","updated_at":"2026-10-15T17:25:09Z","url":"https://research.wusong.net/cases/transkribus","sources":[{"label":"Reviewed primary source: Transkribus: AI-assisted transcription for archival research","url":"https://www.transkribus.org/","kind":"vendor-documentation","is_primary":1}]},{"slug":"tsbench-reaction-mechanisms","title":"TSBench: physics-grounded reaction-mechanism evaluation","summary":"TSBench is a benchmark for evaluating whether LLM agents can construct three-dimensional transition-state guesses consistent with chemical reaction mechanisms.","resource_type":"benchmark-dataset","disciplines":["chemistry"],"tasks":["agent-evaluation","mechanism-elucidation"],"evidence_level":"preprint","outcome_verification":"author-reported","workflow_level":null,"workflow_note":"Resource-level workflow classification unspecified; an instantiated workflow depends on the research task, implementation and oversight. Researchers retain responsibility for validation and safe use.","source_checked_date":"2026-10-17","limitations":"Abstract-only review of arXiv v1. Quantum-chemical settings, benchmark implementation, model configurations and reaction set were not inspected. No downloadable dataset or executable release is confirmed. Automated checking inside the authors' benchmark is not editorial machine-verification or independent reproduction; transfer to synthesis planning is not demonstrated.","adoption_requirements":"Untested editorial guidance: inspect the full benchmark protocol and artifact availability, establish quantum-chemistry expertise and compute requirements, and validate transition-state/reaction-path checks before comparing agents. Check model/data licensing separately.","updated_at":"2026-10-17T15:13:29Z","url":"https://research.wusong.net/cases/tsbench-reaction-mechanisms","sources":[{"label":"TSBench: A physics-grounded benchmark for evaluating LLM understanding of chemical reaction mechanisms — primary arXiv record","url":"https://arxiv.org/abs/2609.08503","kind":"arxiv","is_primary":1}]},{"slug":"university-ai-compute-infrastructure-access","title":"University AI Compute: academic infrastructure and access reference","summary":"Repository-documented university-compute inventory, data pipeline and dashboard relating documented GPU infrastructure to academic research resources and allocation/access regimes. A reference for infrastructure planning and research on access to AI compute, not an automated allocator or a validated performance ranking.","resource_type":"tool","disciplines":["computer-science","education"],"tasks":["data-analysis","data-management"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Infrastructure and access research reference; no AI-executed or autonomous workflow is claimed. Researchers verify source records and institutional access conditions.","source_checked_date":"2026-10-17","limitations":"README-only evidence; underlying CSV records, institutional primary sources, dashboard calculations and validation code were not independently inspected or executed. Inventory is documented lower bounds, not complete capacity: missing/unresolved is not zero, overlapping aggregates must not be double-counted, and generation evidence is sparse. Price-equivalent is a price proxy, not performance equivalence. Correlations are exploratory associations, not causation or evidence of under-investment. GPU counts, node-hours and project counts are different units; cloud/planned systems and merely managed national labs are not owned-GPU totals. No live allocation availability, access entitlement or compute-performance guarantee. Source documents GPU information through October 2026, HERD FY2024 and mixed endowment source years; these are source-described dates, not independently verified currency.","adoption_requirements":"For reference use, read the documented dashboard and methodology; check dated institutional source records and current allocation conditions. Local rebuilding requires inspecting the documented Python pipeline and its data/dependencies first. Data rights, source licences and calculation accuracy need separate review.","updated_at":"2026-10-17T20:39:39Z","url":"https://research.wusong.net/cases/university-ai-compute-infrastructure-access","sources":[{"label":"University AI Compute — README","url":"https://github.com/jryck/university-ai-compute","kind":"code-repository","is_primary":1}]},{"slug":"vibe-coding-review-debugging-comparison-framework","title":"Vibe coding: code-review and debugging comparison framework","summary":"A coding-research paper proposes a secondary-data framework for comparing AI-assisted generation, code review and debugging.","resource_type":"framework","disciplines":["computer-science","education"],"tasks":["experiment-design","research-coding"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level assigned. Selected PDF methods propose a secondary-data evaluation reference: generation speed, integrity and cognitive layers; source/cohort comparability and metric standardisation; prompt-to-code mapping, logic audit and time accounting. A complete AI-assisted research workflow and final human acceptance are not established.","source_checked_date":"2026-10-19","limitations":"Primary abstract, selected publisher PDF methods/limitations/reference context and stated journal policy only; not a full critical audit of data, code or tools. The abstract’s 50-developer wording differs from the body’s secondary-data/meta-analytical virtual experiment and minimum/aggregate cohort rules; no new 50-participant experiment is claimed. The parsed PDF may be cached; raw/version parity is unverified. Prestige/source filters do not prove quality, and operational metrics and included-data traceability remain unclear. Strong time, bug-rate, causal, psychological and pedagogical effects were not validated. Tangential references do not establish misconduct. The source label is not an audit of review conduct, indexing or impact factor.","adoption_requirements":"Untested suggestion: inspect protocol, task complexity, metrics, cohort selection and baseline comparability before designing an evaluation.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/vibe-coding-review-debugging-comparison-framework","sources":[{"label":"Primary article abstract","url":"https://doi.org/10.36948/ijfmr.2026.v08i04.85314","kind":"doi","is_primary":1},{"label":"Selected publisher PDF methods and limitations","url":"https://www.ijfmr.com/papers/2026/4/85314.pdf","kind":"doi","is_primary":1}]},{"slug":"vibe-coding-voice-code-review-provotype","title":"Vibe coding a qualitative research provotype","summary":"A work-in-progress methodological proposal uses AI-assisted coding to build a voice-based code-review research probe, intended to elicit qualitative reactions from Danish software developers.","resource_type":"workflow-process","disciplines":["computer-science","social-science"],"tasks":["experiment-design","research-coding"],"evidence_level":"project-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Researcher-led workflow: people set research questions, approve data access and adjudicate outputs; no autonomous end-to-end performance claim. Workflow level intentionally unspecified.","source_checked_date":"2026-10-17","limitations":"Abstract-only publisher review. Work-in-progress paper; the source explicitly leaves finalizing the prototype, research design and data collection as next steps. No evaluated study outcome, canonical software distribution, runnable instructions or independent reproduction confirmed. Evidence label project-documented is conservative: proceedings presence alone does not establish this work-in-progress section underwent peer review.","adoption_requirements":"Researcher coding oversight, suitable AI coding tools, secure treatment of any developer code, participant consent and qualitative research design. No installable artifact verified.","updated_at":"2026-10-17T19:24:37Z","url":"https://research.wusong.net/cases/vibe-coding-voice-code-review-provotype","sources":[{"label":"Vibe Coding a Research Probe for Exploring AI/Voice Based Code Reviews","url":"https://doi.org/10.34190/icair.5.1.3975","kind":"doi","is_primary":1}]},{"slug":"vision-guided-mobile-manipulator-sdl","title":"Vision-guided mobile manipulation for self-driving laboratories","summary":"A laboratory mobile-manipulator workflow combining kinematic modelling with vision-based detection and pose estimation to support grasping and object-following tasks.","resource_type":"workflow-process","disciplines":["chemistry","robotics"],"tasks":["experiment-execution","image-analysis"],"evidence_level":"peer-reviewed","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Workflow level intentionally unspecified for a reusable component or conceptual reference; researcher oversight and task-specific validation remain necessary.","source_checked_date":"2026-10-17","limitations":"Abstract-only. This is machine-vision-assisted laboratory handling, not evidence of a trained LLM or learned controller. No vision code, calibration procedure, hardware, quantitative accuracy, safety evaluation or experiment was inspected. Claimed scalability and reproducibility benefits are not established by this review. Additional Springer journal abstract and publication citation were inspected after the independent audit; the full journal article was not read. Journal metadata establishes publication class, not independent outcome verification.","adoption_requirements":"Requires a compatible manipulator, camera/depth sensing, kinematic calibration and laboratory safety controls; exact hardware and setup instructions were not inspected.","updated_at":"2026-10-17T15:18:43Z","url":"https://research.wusong.net/cases/vision-guided-mobile-manipulator-sdl","sources":[{"label":"Kinematic Analysis and Integration of Vision Algorithms for a Mobile Manipulator Employed Inside a Self-Driving Laboratory","url":"https://arxiv.org/abs/2510.19081","kind":"arxiv","is_primary":1},{"label":"Official Springer journal abstract and article citation","url":"https://doi.org/10.1007/s41315-025-00518-3","kind":"doi","is_primary":1}]},{"slug":"wealth-capital-consequence-evidence","title":"WEALTH: advisory capital-consequence research infrastructure","summary":"A documented capital-analysis evidence component models consequences and returns advisory evidence through a governance path ending in human decisions.","resource_type":"tool","disciplines":["business","economics"],"tasks":["data-analysis","knowledge-synthesis"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Intended L2 for bounded capital-consequence analysis, source/freshness-qualified advisory evidence and a human decision endpoint. Authority enforcement, model accuracy and handoff were not audited. Forecast calibration/historical replay are planned, not demonstrated L4, investment advice or prediction advantage.","source_checked_date":"2026-10-19","limitations":"Full README only; linked architecture/roadmap, endpoints, models, code and market data not inspected or executed. Source-reported dated healthy/sealed status is not fresh verification. README itself discloses17declared versus18observed tools, unmeasured handoff/freshness coverage, single-reviewer/succession limits and no predictive superiority/external benchmark. Planned calibration/replay is notL4 execution or verified forecasting. No financial advice, capital allocation, transaction or provider connection produced. AI editorial review only; no source code, models or installation executed, no production/status/publication changes.","adoption_requirements":"Read documentation and establish independent model/data rights, governance and provider review before any separate trial. No connector/endpoint/configuration used; README quick-start is not an executed recipe.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/wealth-capital-consequence-evidence","sources":[{"label":"ariffazil/WEALTH primary-readme","url":"https://raw.githubusercontent.com/ariffazil/WEALTH/dde559c8f863ede8ed7e35919dd16215146e0645/README.md","kind":"code-repository","is_primary":1}]},{"slug":"wenyuchiou-ai-research-skills-catalog","title":"AI Research Skills — WenyuChiou catalog","summary":"A catalog of portable research skills covering literature work, research design, durable project context, manuscript work and cross-agent delegation. The README distinguishes this registry from separately maintained skill repositories and the research-hub runtime.","resource_type":"agent-skill","disciplines":["multidisciplinary"],"tasks":["literature-research","scientific-writing","workflow-orchestration"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation is unspecified; researchers select tasks, review sources, validate outputs and authorize consequential actions.","source_checked_date":"2026-10-17","limitations":"README-only evidence; no plugin install, inference, runtime or cross-host compatibility test performed. Maintainer timing and portability figures are not independently verified. The catalog notes domain bias and lack of corpus-scale validation; compatibility depends on the host. The README describes portable skills here but later calls this a registry, not a monorepo; verify each separate upstream source rather than assuming all runnable skills reside in this tree.","adoption_requirements":"A compatible skill-loading AI host; optional runtime and library integrations have separate dependencies. Review prompts, permissions, confidential manuscripts and data rights before use; back up libraries before write-capable integrations.","updated_at":"2026-10-17T20:06:54Z","url":"https://research.wusong.net/cases/wenyuchiou-ai-research-skills-catalog","sources":[{"label":"AI Research Skills — WenyuChiou catalog","url":"https://github.com/WenyuChiou/ai-research-skills","kind":"code-repository","is_primary":1}]},{"slug":"whats-up-timeseries","title":"Whats Up Timeseries: ML examples and reading notes","summary":"Personal collection of time-series/deep-learning reading notes and small ML examples, including ReLU image fitting and a gin-config parameter-configuration notebook.","resource_type":"tutorial-template","disciplines":["computer-science"],"tasks":["research-coding"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"General tutorial collection; automation level unspecified. Researchers select and validate examples, inspect dependencies and judge applicability to their own experiments.","source_checked_date":"2026-10-17","limitations":"README and example-documentation review only; notebooks and wiki articles were not executed or comprehensively reviewed. This is a personal tutorial collection, not an AI-generated literature review or an evaluated forecasting tool. No research-performance gains verified.","adoption_requirements":"Editorial, untested guidance: inspect notebook imports and supply a compatible Python/notebook environment plus dependencies; gin-config is the demonstrated configuration library. No complete environment specification verified.","updated_at":"2026-10-17T14:30:44Z","url":"https://research.wusong.net/cases/whats-up-timeseries","sources":[{"label":"Whats Up Timeseries repository README","url":"https://gitlab.com/Emmanuel_B/whats-up-timeseries","kind":"code-repository","is_primary":1},{"label":"Gin Hello World example README","url":"https://gitlab.com/api/v4/projects/17936914/repository/files/notebooks%2Fhello_gin%2FREADME.md/raw?ref=master","kind":"code-repository","is_primary":1}]},{"slug":"whisperx-research-notebook","title":"WhisperX transcription notebook for qualitative researchers","summary":"A researcher-oriented notebook template for batch transcription, speaker diarization, timestamps and export to analysis-friendly formats.","resource_type":"tutorial-template","disciplines":["humanities","social-science"],"tasks":["data-analysis","transcription"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level entry: no universal autonomy level is assigned. The researcher defines the task, controls access and validates outputs; classify an actual instantiated workflow separately.","source_checked_date":"2026-10-15","limitations":"The notebook was not executed or security-audited. Speech recognition, speaker attribution and pseudonym replacement can fail; privacy and consent must be checked before processing interviews. Pinned upstream versions may be stale.","adoption_requirements":"Notebook/Python runtime, media files, FFmpeg and model dependencies; the README requires a Hugging Face token for diarization. Public repository does not by itself establish reuse permission.","updated_at":"2026-10-15T17:25:09Z","url":"https://research.wusong.net/cases/whisperx-research-notebook","sources":[{"label":"Reviewed primary source: WhisperX transcription notebook for qualitative researchers","url":"https://github.com/mrhallonline/WhisperXTranscription4Researchers","kind":"code-repository","is_primary":1}]},{"slug":"wikistub-seed","title":"WikiStub-Seed: multilingual research-context seed framework","summary":"A documented multilingual JSON knowledge-stub framework supports researcher concept curation, structured exports and optional AI/LLM context or translation integrations.","resource_type":"framework","disciplines":["multidisciplinary"],"tasks":["data-management","knowledge-synthesis"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level assigned: concept seeds, structure checks, exports and optional AI-context/translation integrations are documented, but a concrete supervised AI research method and semantic output-validation protocol are not. Structural validity is not scientific or multilingual correctness.","source_checked_date":"2026-10-19","limitations":"Selected README identity/features/use cases, data-flow/personas, runtime caveats, schema, installation/translation budget and PWA/licence/test notes read; dataset, all siblings, code/API docs not audited. Counts/verified-terminology and tests are source claims, not independently validated research benefit. English relevance slots are currently empty and use German fallback per README; six dataset languages do not mean whole UI translation. Optional translation sends selected text/instructions to Anthropic; network depends on commands, so no general zero-egress claim. PWA offline availability requires prior successful fetch; response targets are not service guarantees. AI editorial review only; no source code, models or installation executed, no production/status/publication changes.","adoption_requirements":"README describes Python source-checkout core, optional integrations and an optional loopback editor; inspect data definitions/rights and command-dependent network/budget controls before any separate trial. No installation, local server, API or dataset processing performed.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/wikistub-seed","sources":[{"label":"dev-bricks/WikiStub-Seed primary-readme","url":"https://raw.githubusercontent.com/dev-bricks/WikiStub-Seed/c1a78b49689ad05a06c67e54afce7188b01cd9ec/README.md","kind":"code-repository","is_primary":1}]},{"slug":"world-cup-2026-baseline-model-research","title":"World Cup 2026 AI Quant: baseline prediction and evaluation workflow","summary":"A documented educational football-research workflow uses transparent predictive baselines, feature experiments and explicit evaluation records.","resource_type":"use-case","disciplines":["computer-science","social-science"],"tasks":["data-analysis","experiment-design"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":"L2","workflow_note":"Intended L2 for analyst-orchestrated predictive research: validate data, choose baseline features, train statistical/ML models, evaluate splits, calibration and uncertainty, and report limitations. Non-agentic modelling does not establish L3/L4, prediction advantage or evaluated tournament outcomes.","source_checked_date":"2026-10-19","limitations":"README and short methodology note only. Data, notebooks, source modules, tests and external resource portals were not inspected or executed. Structure, licensing and CI statements are documentation claims. No complex-model result, calibrated forecast, betting or financial recommendation is established. Moving HEAD documents are not commit-pinned or an atomic version. Promotional links are not endorsed.","adoption_requirements":"Untested suggestion: choose a legitimate dataset, record targets, time splits and baseline metrics, and check leakage and uncertainty before an experiment.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/world-cup-2026-baseline-model-research","sources":[{"label":"Project README snapshot","url":"https://raw.githubusercontent.com/buivanthienhb87-spec/world-cup-2026-ai-quant/HEAD/README.md","kind":"code-repository","is_primary":1},{"label":"Methodology note snapshot","url":"https://raw.githubusercontent.com/buivanthienhb87-spec/world-cup-2026-ai-quant/HEAD/docs/methodology.md","kind":"project-documentation","is_primary":1}]},{"slug":"write-science-well","title":"write-science-well: author-directed scientific prose revision","summary":"An open-format agent skill for researcher-directed scientific prose work. The assessed use revises a supplied passage while retaining its claims, figures, citations and terminology, with proposed edits presented for author acceptance.","resource_type":"agent-skill","disciplines":["multidisciplinary"],"tasks":["scientific-writing"],"evidence_level":"repository-documented","outcome_verification":"author-reported","workflow_level":"L1","workflow_note":"L1 for one researcher-supplied scientific passage revision. The author defines scope and checks/accepts reversible edits while preserving claims, numbers and citations. Internal editing/scanner rounds do not establish autonomous research; small model-graded smoke tests are author-reported.","source_checked_date":"2026-10-18","limitations":"Documentation review of the pinned README, complete core SKILL and house-style only, not every task-specific reference file. A prohibition on fabrication is an instruction, not a guarantee. English sentence-level rules/scanner are not validated multilingual editing. Maintainer evaluation is small/model-graded; no skill/scanner or manuscript experiment was run.","adoption_requirements":"A compatible host assistant or reader of the open SKILL format, author-supplied text/evidence and explicit editing constraints. Protect confidential manuscripts under the chosen host's policies; author verification remains necessary. The scanner needs a suitable Python environment if used; this review did not install or run it.","updated_at":"2026-10-19T20:14:46Z","url":"https://research.wusong.net/cases/write-science-well","sources":[{"label":"README identifies a scientific-writing skill, modes, prerequisites, maintainer smoke-test limits and stated licensing.","url":"https://github.com/geointelmining848/write-science-well","kind":"code-repository","is_primary":1},{"label":"Core instructions define author-directed editing, protected factual content, staged revision and advisory signals.","url":"https://raw.githubusercontent.com/geointelmining848/write-science-well/9b5b7440d6632b3626aa161753d301b1579ac65e/SKILL.md","kind":"project-documentation","is_primary":1},{"label":"House style specifies reversible proposals with author acceptance/refusal and bounded edit scope.","url":"https://raw.githubusercontent.com/geointelmining848/write-science-well/9b5b7440d6632b3626aa161753d301b1579ac65e/references/house-style.md","kind":"project-documentation","is_primary":1}]},{"slug":"writefull","title":"Writefull","summary":"Writefull offers AI language feedback, paraphrasing and copyediting tailored to academic writing.","resource_type":"tool","disciplines":["multidisciplinary"],"tasks":["scientific-writing"],"evidence_level":"vendor-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans define objectives, control data access, and check outputs.","source_checked_date":"2026-10-15","limitations":"No independent language-quality evaluation or verification of privacy promises. Researchers must preserve meaning, verify generated abstracts and comply with disclosure policies.","adoption_requirements":"Access to Writefull and a manuscript the researcher is permitted to process with the service.","updated_at":"2026-10-16T14:58:30Z","url":"https://research.wusong.net/cases/writefull","sources":[{"label":"Writefull","url":"https://www.writefull.com","kind":"vendor-documentation","is_primary":1}]},{"slug":"xaixarts4-proceedings-reference-index","title":"XAIxArts4: proceedings reference index for explainable-AI arts research","summary":"The original proceedings index provides source-discovery pointers for explainable-AI arts research.","resource_type":"framework","disciplines":["arts","computer-science"],"tasks":["knowledge-synthesis","literature-research"],"evidence_level":"preprint","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"No workflow level assigned. The proceedings index provides titles and report-number discovery links across explainable-AI arts research. It is a collection/index, not a single implemented XAI method; constituent AI actions, human oversight and output validation were not assessed.","source_checked_date":"2026-10-19","limitations":"Short official arXiv HTML index and abstract only. Individual papers, workshop website, ACM record and external search results were not audited. Submission acceptance, workshop date/affiliation and the ten-entry listing are source statements, not verified peer-review processes or benefits. The planned resource hub is not assumed delivered. Linked papers and the archived workshop description are distinct resources, not substitutes for the proceedings identity.","adoption_requirements":"Untested suggestion: follow a specific paper’s primary record and inspect its methods, limitations and reuse rights before scholarly use.","updated_at":"2026-10-19T21:37:59Z","url":"https://research.wusong.net/cases/xaixarts4-proceedings-reference-index","sources":[{"label":"Official proceedings index","url":"https://arxiv.org/html/2607.20131","kind":"arxiv","is_primary":1}]},{"slug":"yliu623-ai-research-digest","title":"AI Research Digest","summary":"Repository-documented workflow using a Claude Code cloud routine to prepare daily AI/LLM research and news digests for economics, finance and operations, with a searchable archive and an audio-briefing build.","resource_type":"workflow-process","disciplines":["business","economics"],"tasks":["knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level workflow level unspecified; documented configuration is not verified autonomous operation. Humans control scope and inspect outputs.","source_checked_date":"2026-10-17","limitations":"README-only review; PROMPT.md, source-fetching code, generated digests and audio were not audited or executed. Completeness, factual accuracy, delivery reliability and source permissions remain unverified.","adoption_requirements":"Claude Code cloud routine configuration; repository and Python build environment; Gemini API credentials for the documented audio step. Check provider costs, source access rights and recipient consent; keep credentials and unpublished material private.","updated_at":"2026-10-17T17:46:50Z","url":"https://research.wusong.net/cases/yliu623-ai-research-digest","sources":[{"label":"AI Research Digest — primary README","url":"https://raw.githubusercontent.com/yliu623/ai-research-digest/HEAD/README.md","kind":"code-repository","is_primary":1}]},{"slug":"zetetic-team-subagents","title":"Zetetic Agents: evidence-synthesis and research reasoning skills","summary":"Repository-documented agent skills for evidence synthesis and research reasoning. The README describes portable evidence-synthesis packaging for Codex and Gemini CLI, alongside a fuller Claude Code distribution and host-specific quality gates.","resource_type":"agent-skill","disciplines":["multidisciplinary"],"tasks":["knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation unspecified. The researcher defines the task, controls access and checks sources and outputs.","source_checked_date":"2026-10-17","limitations":"README-only review of a mutable main branch; no installation or research outcome reproduced. Citation-presence checks do not establish citation validity; documented refusals are prompt guidance rather than runtime guarantees. Hooks can be bypassed outside supported host invocation paths. Companion Cortex and session-optimizer projects are distinct, not aliases.","adoption_requirements":"Requires a compatible agent host; portable and full distributions differ. Inspect hook permissions and source-data policies before installation; retain researcher judgment over causal claims and evidence quality.","updated_at":"2026-10-17T17:46:50Z","url":"https://research.wusong.net/cases/zetetic-team-subagents","sources":[{"label":"cdeust/zetetic-team-subagents — primary README","url":"https://github.com/cdeust/zetetic-team-subagents","kind":"code-repository","is_primary":1}]},{"slug":"zotero-llm-wiki-paper-chat","title":"zotero-llm-wiki: Paper Chat and literature skills","summary":"Repository-documented Zotero and Claude tools: Paper Chat provides PDF questions with linked quotations and highlights; paper-ask supports library questions; paper-search uses OpenAlex; the full Claude Code plugin maintains an Obsidian literature database. These are documented capabilities, not guarantees of accurate citation or exhaustive retrieval.","resource_type":"tool","disciplines":["multidisciplinary"],"tasks":["data-management","knowledge-synthesis","literature-research"],"evidence_level":"repository-documented","outcome_verification":"not-evaluated","workflow_level":null,"workflow_note":"Resource-level automation unspecified; researchers define aims and verify outputs. No autonomous end-to-end outcome claimed.","source_checked_date":"2026-10-17","limitations":"README-only substantive review plus repository identity metadata; no installation or outcome tests. Paper Chat sends the whole PDF text to Anthropic, cannot see figures and needs OCR for scanned PDFs. The README says Paper Chat was tested on macOS but not Windows; this is the author’s statement. Full-version generated fields are currently Chinese, and rebuilding generated pages can overwrite edits. Whole-text transmission applies in either authentication mode: do not use documents that must remain confidential or local. Paper Chat rereads the whole text per question; subscription quotas or per-use API costs apply.","adoption_requirements":"Paper Chat requires Zotero and Claude Code or an Anthropic API key. Skills require Claude Code and Python; the full setup adds Obsidian and Better BibTeX. Review subscription/API costs, document rights, provider transmission and Zotero permissions before adoption.","updated_at":"2026-10-17T20:06:54Z","url":"https://research.wusong.net/cases/zotero-llm-wiki-paper-chat","sources":[{"label":"nebula0/zotero-llm-wiki","url":"https://raw.githubusercontent.com/nebula0/zotero-llm-wiki/main/README.md","kind":"code-repository","is_primary":1}]}],"timestamp_note":"updated_at is the catalogue record edit time; source_checked_date is the recorded source check, not independent reproduction.","evidence_note":"Source level, outcome verification and workflow scope are independent. Inclusion is not endorsement."}