AI tools, agent skills, frameworks and workflows for academic research.
Browse reviewed resources, from everyday literature review, writing and data analysis to
autonomous research. Inclusion is not a claim of proven benefit.
Resources
20 resources in the catalogue.
Workflow level (L1–L4) describes how AI is used, not evidence quality. Higher is not better evidence.
Source level and outcome verification are separate labels. Read the method.
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.
L3Multi-step tool-using agentic workflow (workflow level, not a quality rating)Source: Peer-reviewed sourceOutcome: Machine-verified artifacts
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.
L3Multi-step tool-using agentic workflow (workflow level, not a quality rating)Source: Peer-reviewed sourceOutcome: Author-reported
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.
L4Autonomous iterative research (workflow level, not a quality rating)Source: Peer-reviewed sourceOutcome: Author-reported
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.
L4Autonomous iterative research (workflow level, not a quality rating)Source: Peer-reviewed sourceOutcome: Externally validated (partial)
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.
–Workflow level n/aSource: Repository-documentedOutcome: Not empirically evaluated here
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.
L3Multi-step tool-using agentic workflow (workflow level, not a quality rating)Source: Peer-reviewed sourceOutcome: Machine-verified artifacts
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.
L2Research workflow (workflow level, not a quality rating)Source: Peer-reviewed sourceOutcome: Author-reported
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.
L3Multi-step tool-using agentic workflow (workflow level, not a quality rating)Source: Preprint (not peer-reviewed)Outcome: Author-reported
3 sourcesSource-checked 2026-10-15BiologyMultidisciplinaryKnowledge synthesisLiterature research
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.
–Workflow level n/aSource: Mixed sourcesOutcome: Not empirically evaluated here
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.
L4Autonomous iterative research (workflow level, not a quality rating)Source: Peer-reviewed sourceOutcome: Author-reported
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.
L4Autonomous iterative research (workflow level, not a quality rating)Source: Peer-reviewed sourceOutcome: Author-reported
An open-source machine-learning pipeline helps researchers screen titles and abstracts for systematic reviews and meta-analyses, with simulation and user-experience evaluations.
L2Research workflow (workflow level, not a quality rating)Source: Peer-reviewed sourceOutcome: Author-reported
2 sourcesSource-checked 2026-10-15MultidisciplinaryKnowledge synthesisLiterature research
Use caseStringer, Wang, Michaelos and Pachitariu / MouseLand2021
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.
L1Task assistance (workflow level, not a quality rating)Source: Peer-reviewed sourceOutcome: Author-reported