eGoT: enhanced graph-of-thoughts for multi-hop knowledge retrieval and hypothesis generation in biomedicine.

eGoT achieves superior multi-hop reasoning in biomedical question answering, outperforming state-of-the-art methods on multiple datasets with 1046 PubMed articles.

• Why it matters: Understanding complex biological mechanisms and disease processes requires integrating fragmented evidence from unstructured literature, which current models struggle to do accurately and transparently.
• What they did: The approach combines automated knowledge graph construction from biomedical texts with a novel graph-of-thoughts querying method, enabling multi-round LLM-based retrieval across multiple domains.
• The result: eGoT enables comprehensive, provenance-backed responses to complex biomedical questions, demonstrated through case studies on lung cancer and lupus, advancing hypothesis generation and knowledge integration.