The abstract is read at the publisher; the summary is JClub's.
statLens delivers 92.5–95.0% accuracy on 102 benchmark differential expression tasks across 74 biomedical omics datasets, outperforming LLM-only inference (17.5–46.8%) across seven language models.
- Why it matters: Large-scale omics workflows are complex and hard to apply directly with LLMs, and existing methods lack integrated accuracy and interpretability with reliable human oversight.
- What they did: The framework combines LLMs, generated code, and human review, evaluated on 102 benchmark tasks spanning six omics modalities and compared against LLM-only and autonomous-agent approaches.
- The result: Decomposing workflows into code, LLM, and expert steps yields consistently high accuracy and improved interpretability and reproducibility, including gains for four small language models versus autonomous agents and LLM-only inference.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.