The abstract is read at the publisher; the summary is JClub's.
Interaction-finder recovers up to 4.3× as many known biological entity associations across 60 literature topics, while verifying every extracted quote against source text.
- Why it matters: Curating interaction lists is slow when no database covers a given question, forcing researchers to manually scan papers.
- What they did: The tool uses LLM-guided iterative literature search, extracts candidate associations from full-text, and outputs a ranked HTML report with quote-level provenance verified against the source.
- The result: Gold-standard associations are enriched near the top, achieving overall recall@20 of 0.61, and unverified candidates scored by a blind LLM judge similarly to gold-standard entries.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.