Literature Meets Data (hypothesis-generation framework)

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]

Workflow level n/a Source Peer-reviewed source Outcome Author-reported

At a glance

Type

Framework

Workflow level

Workflow unspecified / not applicable

Organisation

University of Chicago

Year

2025

Disciplines

Computer scienceSocial science

Tasks

Hypothesis generationKnowledge synthesis

Entry updated

2026-10-17

Evidence & verification

Source level

Peer-reviewed source
Grounded in a peer-reviewed publication. This describes the source; it does not mean the reported results have been independently reproduced.

Outcome verification

Author-reported
Outcomes are reported by the team that built the system. The entry is source-checked against the team's own publications; not independently reproduced.

Sources checked

2026-10-16 (UTC)

Authors' own publications (preprint and ACL 2025 records checked live 2026-10-16).[14] Performance claims are not independently reproduced.

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

Sources (1)

Research applicability

Getting started

Review the paper (arXiv:2410.17309; ACL 2025 version DOI 10.18653/v1/2025.acl-long.12); availability not verified.

Maintenance

Maintenance status not independently audited.

License

Not independently audited.

Workflow & human oversight

Resource-level classification is intentionally unspecified; instantiated workflows may differ. Humans set objectives and validate results.

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

Approach

First method reported to combine literature-derived insights and dataset-grounded analysis for LLM hypothesis generation; evaluated on five datasets plus a human evaluation assessing LLM-generated hypotheses as decision aids (deception detection; AI-generated content detection).[14]

Outcomes (as reported by the originating team)

Author-reported: 8.97% over few-shot, 15.75% over literature-based alone and 3.37% over data-driven alone; human accuracy improved by 7.44% and 14.19% on the two human-evaluation tasks.

Limitations

Author-reported results; not independently reproduced (including the human evaluation). Published version checked at record/abstract level.

Reproducibility

Live source checks 2026-10-16: arXiv:2410.17309; ACL 2025 DOI 10.18653/v1/2025.acl-long.12. No claimed outcome reproduced.

Adoption requirements

LLM access; literature corpus and task datasets; no packaged implementation verified.

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.