A New Role for Relevance: Guiding Corpus Interaction in Agentic Search
Jiangnan Li, Yuqing Li, Mo Yu, Jinchao Zhang et al. · arXiv · 2026
Abstract
Retrieval agents normally use relevance scores to pick a top-k set and stop there. This work uses relevance differently — as an ordering prior for how an agent explores a corpus with grep-style search: which documents to walk first, where to begin reading inside them, and which matches to surface ahead of the rest. Questions whose evidence must be located and assembled, rather than simply fetched, converge faster as a result.
Why it matters
Worth reading if you build RAG and keep hitting the ceiling of top-k retrieval. It reframes ranking as something that steers the search process rather than something that filters the final context window.
https://arxiv.org/abs/2607.24223