{"paper":{"title":"CogRAG: Tackling Heterogeneous Cognitive Demands in RAG via Stratified Retrieval and Reasoning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"CogRAG+ separates retrieval from reasoning in LLMs using dual paths and structured templates to fix knowledge gaps on professional exams.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Kui Su, Xudong Wang, Zhaoyan Ming, Zilong Wang","submitted_at":"2026-04-01T13:31:12Z","abstract_excerpt":"Retrieval-Augmented Generation (RAG) frameworks typically process all queries through a one-size-fits-all pipeline, ignoring the heterogeneous cognitive demands of different tasks. This cognitive-blind approach causes two failure modes: cascading errors when low-level factual gaps trigger hallucinated reasoning, and reasoning-answer inconsistency in higher-order analytical tasks. We introduce CogRAG, a training-free, domain-agnostic framework that tackles these heterogeneous cognitive demands via stratified retrieval and reasoning. Inspired by Bloom's Taxonomy, CogRAG uses the predicted cognit"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Experiments on two representative models, Qwen3-8B and Llama3.1-8B, show that CogRAG+ consistently outperforms general-purpose models and standard RAG methods on the Registered Dietitian qualification exam. In single-question mode, it raises overall accuracy to 85.8% for Qwen3-8B and 60.3% for Llama3.1-8B, with clear gains over vanilla baselines.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumption that a judge-driven dual-path retrieval strategy can reliably identify and supply missing foundational knowledge without domain-specific tuning or additional training data.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"CogRAG+ raises LLM accuracy on a dietitian exam to 85.8% by using dual-path retrieval and structured reasoning templates.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"CogRAG+ separates retrieval from reasoning in LLMs using dual paths and structured templates to fix knowledge gaps on professional exams.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"31c80d4001398e2cebb6dd379eeb4b14d9922f9e5d72a8c5e52c55572ae9d3e6"},"source":{"id":"2604.25928","kind":"arxiv","version":2},"verdict":{"id":"cda35c8d-83bd-4227-b831-0118372c87fa","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-13T22:25:57.600245Z","strongest_claim":"Experiments on two representative models, Qwen3-8B and Llama3.1-8B, show that CogRAG+ consistently outperforms general-purpose models and standard RAG methods on the Registered Dietitian qualification exam. In single-question mode, it raises overall accuracy to 85.8% for Qwen3-8B and 60.3% for Llama3.1-8B, with clear gains over vanilla baselines.","one_line_summary":"CogRAG+ raises LLM accuracy on a dietitian exam to 85.8% by using dual-path retrieval and structured reasoning templates.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumption that a judge-driven dual-path retrieval strategy can reliably identify and supply missing foundational knowledge without domain-specific tuning or additional training data.","pith_extraction_headline":"CogRAG+ separates retrieval from reasoning in LLMs using dual paths and structured templates to fix knowledge gaps on professional exams."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.25928/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":1,"snapshot_sha256":"5ae7427c5dec90fa7a133fe61c0eb10fc6d6deb82aacadf975738200f7c5b8b6"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}