{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:L7BQSG7C6IDX7FF6WVVLL5ABSD","short_pith_number":"pith:L7BQSG7C","schema_version":"1.0","canonical_sha256":"5fc3091be2f2077f94beb56ab5f40190fc0594234619ca919a10d79f135f0d34","source":{"kind":"arxiv","id":"2504.00374","version":1},"attestation_state":"computed","paper":{"title":"When Persuasion Overrides Truth in Multi-Agent LLM Debates: Introducing a Confidence-Weighted Persuasion Override Rate (CW-POR)","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Divyam Khanna, Mahak Agarwal","submitted_at":"2025-04-01T02:45:02Z","abstract_excerpt":"In many real-world scenarios, a single Large Language Model (LLM) may encounter contradictory claims-some accurate, others forcefully incorrect-and must judge which is true. We investigate this risk in a single-turn, multi-agent debate framework: one LLM-based agent provides a factual answer from TruthfulQA, another vigorously defends a falsehood, and the same LLM architecture serves as judge. We introduce the Confidence-Weighted Persuasion Override Rate (CW-POR), which captures not only how often the judge is deceived but also how strongly it believes the incorrect choice. Our experiments on "},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2504.00374","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-04-01T02:45:02Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"ff08d4eb5484cc7731ed650516cbe9453a42e0ac252bc12ecf812fdf1cd6cac2","abstract_canon_sha256":"03f9401ed6fb71bab4767bd09e3d4329efc8874551d1074dc3618b1077f2c71c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:42:25.218067Z","signature_b64":"Hfzb0prMKsYexTk02JukGbIutpcpssxFHy3wrqpQAeKIEjelqm4P40oee6rwvwryl3od863h2j42RxA3gRxHAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5fc3091be2f2077f94beb56ab5f40190fc0594234619ca919a10d79f135f0d34","last_reissued_at":"2026-07-05T10:42:25.217604Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:42:25.217604Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"When Persuasion Overrides Truth in Multi-Agent LLM Debates: Introducing a Confidence-Weighted Persuasion Override Rate (CW-POR)","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Divyam Khanna, Mahak Agarwal","submitted_at":"2025-04-01T02:45:02Z","abstract_excerpt":"In many real-world scenarios, a single Large Language Model (LLM) may encounter contradictory claims-some accurate, others forcefully incorrect-and must judge which is true. We investigate this risk in a single-turn, multi-agent debate framework: one LLM-based agent provides a factual answer from TruthfulQA, another vigorously defends a falsehood, and the same LLM architecture serves as judge. We introduce the Confidence-Weighted Persuasion Override Rate (CW-POR), which captures not only how often the judge is deceived but also how strongly it believes the incorrect choice. Our experiments on "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.00374","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2504.00374/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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2504.00374","created_at":"2026-07-05T10:42:25.217659+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.00374v1","created_at":"2026-07-05T10:42:25.217659+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.00374","created_at":"2026-07-05T10:42:25.217659+00:00"},{"alias_kind":"pith_short_12","alias_value":"L7BQSG7C6IDX","created_at":"2026-07-05T10:42:25.217659+00:00"},{"alias_kind":"pith_short_16","alias_value":"L7BQSG7C6IDX7FF6","created_at":"2026-07-05T10:42:25.217659+00:00"},{"alias_kind":"pith_short_8","alias_value":"L7BQSG7C","created_at":"2026-07-05T10:42:25.217659+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2510.07517","citing_title":"When Identity Skews Debate: Anonymization for Bias-Reduced Multi-Agent Reasoning","ref_index":56,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/L7BQSG7C6IDX7FF6WVVLL5ABSD","json":"https://pith.science/pith/L7BQSG7C6IDX7FF6WVVLL5ABSD.json","graph_json":"https://pith.science/api/pith-number/L7BQSG7C6IDX7FF6WVVLL5ABSD/graph.json","events_json":"https://pith.science/api/pith-number/L7BQSG7C6IDX7FF6WVVLL5ABSD/events.json","paper":"https://pith.science/paper/L7BQSG7C"},"agent_actions":{"view_html":"https://pith.science/pith/L7BQSG7C6IDX7FF6WVVLL5ABSD","download_json":"https://pith.science/pith/L7BQSG7C6IDX7FF6WVVLL5ABSD.json","view_paper":"https://pith.science/paper/L7BQSG7C","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.00374&json=true","fetch_graph":"https://pith.science/api/pith-number/L7BQSG7C6IDX7FF6WVVLL5ABSD/graph.json","fetch_events":"https://pith.science/api/pith-number/L7BQSG7C6IDX7FF6WVVLL5ABSD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/L7BQSG7C6IDX7FF6WVVLL5ABSD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/L7BQSG7C6IDX7FF6WVVLL5ABSD/action/storage_attestation","attest_author":"https://pith.science/pith/L7BQSG7C6IDX7FF6WVVLL5ABSD/action/author_attestation","sign_citation":"https://pith.science/pith/L7BQSG7C6IDX7FF6WVVLL5ABSD/action/citation_signature","submit_replication":"https://pith.science/pith/L7BQSG7C6IDX7FF6WVVLL5ABSD/action/replication_record"}},"created_at":"2026-07-05T10:42:25.217659+00:00","updated_at":"2026-07-05T10:42:25.217659+00:00"}