BioMedAgent: Nature-Published Multi-Agent Framework Achieves 77% on 327 Biomedical Data Tasks

Nature Biomedical Engineering·medium signalXBlueskyLinkedInCopy link

Published in Nature Biomedical Engineering, BioMedAgent is a self-evolving LLM multi-agent framework that learns to use bioinformatics tools and chain them into executable workflows from natural-language prompts. Evaluated on the BioMed-AQA benchmark (327 tasks), it achieved 77% success rate — surpassing all other LLM agents — and generalized to the external BixBench dataset. It autonomously performs cross-omics analysis, ML modeling, and pathology image segmentation without requiring computational expertise from users.