CORE improves search accuracy by managing context and snippet length
CORE: Coarse-to-Fine Search with Context Management for Deep Research
Computation and LanguageArtificial IntelligenceInformation Retrieval
Summary
When computers try to answer tough questions, they search through lots of information. Finding just the right bits is tricky because the computer needs enough context to understand the information but also can’t remember too much at once. The authors made CORE, a system that first looks broadly for useful passages and then zooms in to find the most important snippet. CORE also keeps track of what it has already seen to avoid repeating the same facts. It works better than previous methods at giving correct answers and finding the right evidence.
What this means in practice
- •For search engine developers: Enhance search engines to return concise, relevant snippets that improve answer accuracy without overloading the user's query context.
- •For legal document analysts: Develop tools that retrieve critical legal text passages while managing context to avoid redundancy and improve evidence recall.
Authors
Hoang Tran Vuong, Ngoc Quan Pham, Hung Bui, Fatih Porikli, Dat Quoc Nguyen
Abstract
Deep research agents answer complex questions through repeated search. They rely on a harness that runs their tool calls and returns evidence to their context. This creates a trade-off: Search works best over long passages, which give enough context for reliable semantic matching. But the returned evidence should be short, because every observation stays in the agent's context for the rest of the trajectory. We introduce CORE, a training-free harness that combines coarse-to-fine search with context management. CORE retrieves passages with a dense retriever. It then reuses the token-level similarities from a late-interaction reranker to return only the most relevant snippet from each passage. Context management lets the agent expand the context around sources it has already retrieved and suppresses repeated snippets. In controlled closed-search comparisons with the Tongyi DeepResearch and OpenResearcher harnesses, CORE achieves state-of-the-art answer accuracy and evidence recall across three agent models and three benchmarks (BrowseComp-Plus, BrowseComp-Wikipedia and BrowseComp-Plus-CM), on corpora containing up to 553M documents.