8. MedBrowseComp MedBrowseComp is a new benchmark designed to evaluate LLM agents’ ability to perform complex, multi-hop medical fact-finding by browsing real-world, domain-specific web resources.

→ View original post on X — @dair_ai