New agents benchmark: CollaborativeAgentBench is the first benchmark studying collaborative LLM agents that work with humans across multi-turn collaboration on realistic tasks in backend programming & frontend design

→ View original post on X — @aiatmeta,