BrowseComp+ evidence placement experiments illustrate retrieval bottlenecks in end-to-end question answering

The unreasonable effectiveness of BM25 for agentic search — Jo Kristian Bergum, Hornet.dev

Jo Kristian Bergum

Agentic search as an agent loop combining a capable model, a retrieval harness and an efficient search engine · BM25 lexical scoring and the separate problem of accelerating top-K retrieval · LLM general knowledge and iterative query formulation make lexical search more useful