STORM: a framework for source-grounded pre-writing and knowledge curation

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/a Source Mixed sources Outcome Not empirically evaluated here

At a glance

Type

Framework

Workflow level

Workflow unspecified / not applicable

Organisation

Stanford Open Virtual Assistant Lab

Year

2024

Disciplines

Computer scienceMultidisciplinary

Tasks

Knowledge synthesisLiterature researchScientific writing

Entry updated

2026-10-15

Evidence & verification

Source level

Mixed sources
Grounded in a combination of peer-reviewed and non-peer-reviewed sources; see the individual sources.

Outcome verification

Not empirically evaluated here
A documented resource or intended workflow, not a demonstrated academic outcome. No empirical benefit is established by its inclusion.

Sources checked

2026-10-15 (UTC)

Primary NAACL 2024 publication abstract and repository documentation checked 2026-10-15. The paper evaluates Wikipedia-like article quality; the directory makes no empirical academic-benefit claim for this reusable framework. Repository popularity is not verification.

Source-checked does not mean independently reproduced. See the Method page for how these labels are assigned.

Sources (3)

Research applicability

Can support topic exploration and pre-writing for researchers across disciplines. Its published evaluation concerns Wikipedia-like articles and editor feedback, not academic publication quality or research productivity.

Getting started

Begin with the upstream overview and examples. Decide whether to use a licensed document corpus or a search provider, pin a release, pilot a small question and audit citations before integrating outputs into research writing.

Maintenance

GitHub API checked 2026-10-15: not archived, last pushed_at 2025-09-30. API/service compatibility and unresolved issues were not audited.

License

MIT license verified from the repository LICENSE. Retrieved documents, datasets and model/provider services have separate terms.

Related resources

PaperQA2: language agents achieving superhuman synthesis of scientific knowledge

Workflow & human oversight

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.

Workflow level is not an evidence score; higher is not better evidence.

Problem

Grounded long-form drafting requires researching a topic and organizing sources before composing prose.

Approach

STORM separates pre-writing from writing and exposes configurable language-model and retrieval modules. The README also documents Co-STORM human-AI knowledge curation.

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.

Reproducibility

The README points to the NAACL-2024-code-backup branch for the original STORM experiment. Current main and Co-STORM differ. No code or model was executed for this review.

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.

Comments

No comments yet.

Add a comment

Comments are held for moderation before appearing. Critical or corrective comments are welcome — especially evidence that an entry overstates what has been verified.