LabWeaver: reliable agentic data analysis grounded in laboratory knowled

LabWeaver shows that laboratory-organized experimental context enables correct fusion reporter recovery and improved tumor cell annotation in a synovial sarcoma single-cell RNA-seq dataset.

• Why it matters: Agentic LLM systems can be unreliable because they miss experimental context, and existing designs organize context by users/projects rather than laboratories.
• What they did: LabWeaver uses a three-tier laboratory–user–project architecture, distilling experimental context into shared laboratory memory and skills after human-gated review, then tests it on synovial sarcoma scRNA-seq across two sessions.
• The result: With laboratory memory/skills available, the system recovered reporter expression and resolved five fusion-anchored tumor cell states; without it, it lost the fusion reporter at read alignment and misannotated tumor cells.

AI summary of the abstract · may contain errors · not medical advice · read the paper.