A New Role for Relevance: Guiding Corpus Interaction in Agentic Search
RARG uses relevance as a runtime priority for grep-style agent search, not just as a prefilter.
The paper argues that ordinary document relevance can miss where the useful evidence sits inside a corpus. Its agent orders documents for ripgrep traversal, seeds searches with query-relevant paragraphs, and reranks grep matches so stronger excerpts appear earlier. The authors report better accuracy-efficiency tradeoffs on browse QA and reasoning-heavy retrieval tasks. HF Daily Papers' note
The paper argues that ordinary document relevance can miss where the useful evidence sits inside a corpus. Its agent orders documents for ripgrep traversal, seeds searches with query-relevant paragraphs, and reranks grep matches so stronger excerpts appear earlier. The authors report better accuracy-efficiency tradeoffs on browse QA and reasoning-heavy retrieval tasks. HF Daily Papers' note
score 4