{"paper":{"title":"NaviRAG: Towards Active Knowledge Navigation for Retrieval-Augmented Generation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"NaviRAG replaces flat segment retrieval in RAG with an LLM agent that actively navigates a document hierarchy from coarse topics to fine details.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"2), (2) Nanjing University, (3) Northeastern University), Dingjun Wu (1), Jihao Dai (1, Maosong Sun (1) ((1) Tsinghua University, Yukun Yan (1), Yuxuan Chen (1), Zhenghao Liu (3), Zheni Zeng (2)","submitted_at":"2026-04-14T14:07:01Z","abstract_excerpt":"Retrieval-augmented generation (RAG) typically relies on a flat retrieval paradigm that maps queries directly to static, isolated text segments. This approach struggles with more complex tasks that require the conditional retrieval and dynamic synthesis of information across different levels of granularity (e.g., from broad concepts to specific evidence). To bridge this gap, we introduce NaviRAG, a novel framework that shifts from passive segment retrieval to active knowledge navigation. NaviRAG first structures the knowledge documents into a hierarchical form, preserving semantic relationship"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Extensive experiments on long-document QA benchmarks show that NaviRAG consistently improves both retrieval recall and end-to-end answer performance over conventional RAG baselines.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That automatically structuring documents into a hierarchy will preserve accurate semantic relationships across granularity levels and that an LLM agent can reliably identify information gaps and select the appropriate retrieval level without introducing new errors.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"NaviRAG improves RAG performance on long-document QA by structuring knowledge hierarchically and deploying an LLM agent for iterative, multi-granularity retrieval.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"NaviRAG replaces flat segment retrieval in RAG with an LLM agent that actively navigates a document hierarchy from coarse topics to fine details.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"da7d40e9a17f215c7447d8ae897202250c25c0ae80fd0034050d16ea0eea66c0"},"source":{"id":"2604.12766","kind":"arxiv","version":2},"verdict":{"id":"f0db4315-47be-4e74-829d-d7d1c815a733","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T15:02:13.177825Z","strongest_claim":"Extensive experiments on long-document QA benchmarks show that NaviRAG consistently improves both retrieval recall and end-to-end answer performance over conventional RAG baselines.","one_line_summary":"NaviRAG improves RAG performance on long-document QA by structuring knowledge hierarchically and deploying an LLM agent for iterative, multi-granularity retrieval.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That automatically structuring documents into a hierarchy will preserve accurate semantic relationships across granularity levels and that an LLM agent can reliably identify information gaps and select the appropriate retrieval level without introducing new errors.","pith_extraction_headline":"NaviRAG replaces flat segment retrieval in RAG with an LLM agent that actively navigates a document hierarchy from coarse topics to fine details."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.12766/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}