{"paper":{"title":"Debating the Unspoken: Role-Anchored Multi-Agent Reasoning for Half-Truth Detection","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"Role-anchored multi-agent debate detects half-truths by exposing omitted context in claims.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Anthony K.H. Tung, Hang Feng, Yirui Zhang, Yixuan Tang","submitted_at":"2026-04-21T02:46:10Z","abstract_excerpt":"Half-truths, claims that are factually correct yet misleading due to omitted context, remain a blind spot for fact verification systems focused on explicit falsehoods. Addressing such omission-based manipulation requires reasoning not only about what is said, but also about what is left unsaid. We propose RADAR, a role-anchored multi-agent debate framework for omission-aware fact verification under realistic, noisy retrieval. RADAR assigns complementary roles to a Politician and a Scientist, who reason adversarially over shared retrieved evidence, moderated by a neutral Judge. A dual-threshold"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Experiments show that RADAR consistently outperforms strong single- and multi-agent baselines across datasets and backbones, improving omission detection accuracy while reducing reasoning cost.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That complementary role assignment and adversarial debate over shared retrieved evidence, combined with dual-threshold early termination, reliably uncovers omitted context without introducing new biases or requiring perfect retrieval.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"RADAR is a role-anchored multi-agent debate framework that improves half-truth detection by reasoning about omitted context in noisy retrieved evidence.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Role-anchored multi-agent debate detects half-truths by exposing omitted context in claims.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"a415dafafb805ac6705867db545a83d739cb1adba8b1747f09ab38388a44e005"},"source":{"id":"2604.19005","kind":"arxiv","version":2},"verdict":{"id":"1c31d13b-c1fe-49c8-9cf5-98455a61bc85","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T02:37:38.351275Z","strongest_claim":"Experiments show that RADAR consistently outperforms strong single- and multi-agent baselines across datasets and backbones, improving omission detection accuracy while reducing reasoning cost.","one_line_summary":"RADAR is a role-anchored multi-agent debate framework that improves half-truth detection by reasoning about omitted context in noisy retrieved evidence.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That complementary role assignment and adversarial debate over shared retrieved evidence, combined with dual-threshold early termination, reliably uncovers omitted context without introducing new biases or requiring perfect retrieval.","pith_extraction_headline":"Role-anchored multi-agent debate detects half-truths by exposing omitted context in claims."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.19005/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_compliance","ran_at":"2026-05-20T03:25:15.292392Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"9eb1a1d49ad1fb2d57c4561603c5078d3f7aa78339f099de4e1c6264dfb943b6"},"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"}