{"paper":{"title":"CounterRefine: Answer-Conditioned Counterevidence Retrieval for Inference-Time Knowledge Repair in Factual Question Answering","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"CounterRefine repairs factual answers at inference time by retrieving counterevidence to test and revise provisional responses.","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Tianyi Huang, Ying Kai Deng","submitted_at":"2026-03-17T03:27:25Z","abstract_excerpt":"In factual question answering, many errors are not failures of access but failures of commitment: the system retrieves relevant evidence, yet still settles on the wrong answer. We present CounterRefine, a lightweight repair layer for short-form RAG that treats the first answer as a hypothesis to test. Given a draft, CounterRefine issues answer-conditioned expansion queries to retrieve candidate-specific evidence, then applies a constrained KEEP or REVISE refinement step whose proposed revisions are accepted only after deterministic validation. The design is intentionally narrow: it adds one ev"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"On the full SimpleQA benchmark, CounterRefine improves a matched GPT-5 Baseline-RAG by 5.8 points and reaches a 73.1 percent correct rate, while exceeding the reported one-shot GPT-5.4 score by roughly 40 points.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the restricted refinement step with deterministic validation reliably distinguishes valid revisions from invalid ones across diverse factual questions without introducing new errors.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"CounterRefine improves factual QA by retrieving answer-conditioned counterevidence and deterministically refining draft answers, lifting a GPT-5 RAG baseline by 5.8 points to 73.1% on SimpleQA.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"CounterRefine repairs factual answers at inference time by retrieving counterevidence to test and revise provisional responses.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"1e19c41cf1dc3920a64b6e84b50a0baa39b93e0bd5633e6915f228cc2435f695"},"source":{"id":"2603.16091","kind":"arxiv","version":3},"verdict":{"id":"52c56e36-9719-4641-8649-f5d56d41495f","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T10:39:53.958425Z","strongest_claim":"On the full SimpleQA benchmark, CounterRefine improves a matched GPT-5 Baseline-RAG by 5.8 points and reaches a 73.1 percent correct rate, while exceeding the reported one-shot GPT-5.4 score by roughly 40 points.","one_line_summary":"CounterRefine improves factual QA by retrieving answer-conditioned counterevidence and deterministically refining draft answers, lifting a GPT-5 RAG baseline by 5.8 points to 73.1% on SimpleQA.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the restricted refinement step with deterministic validation reliably distinguishes valid revisions from invalid ones across diverse factual questions without introducing new errors.","pith_extraction_headline":"CounterRefine repairs factual answers at inference time by retrieving counterevidence to test and revise provisional responses."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2603.16091/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"}