{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:SOXYK4SSJ3OVQRK5HZDVYFEEFH","short_pith_number":"pith:SOXYK4SS","schema_version":"1.0","canonical_sha256":"93af8572524edd58455d3e475c148429f48e66f1dbcf24643274a67e6bbba4a4","source":{"kind":"arxiv","id":"2503.11656","version":1},"attestation_state":"computed","paper":{"title":"TRUTH DECAY: Quantifying Multi-Turn Sycophancy in Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Aarav Jain, Aslihan Akalin, Joshua Liu, Kevin Zhu, Sean O'Brien, Soham Takuri, Srihan Vege, Vasu Sharma","submitted_at":"2025-02-04T06:56:32Z","abstract_excerpt":"Rapid improvements in large language models have unveiled a critical challenge in human-AI interaction: sycophancy. In this context, sycophancy refers to the tendency of models to excessively agree with or flatter users, often at the expense of factual accuracy. While previous studies have primarily analyzed this behavior in single-turn interactions, its persistence and evolution in multi-step conversations remain largely unexplored. We introduce TRUTH DECAY, a benchmark specifically designed to evaluate sycophancy in extended dialogues, where language models must navigate iterative user feedb"},"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":"2503.11656","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-02-04T06:56:32Z","cross_cats_sorted":[],"title_canon_sha256":"5706602961198a08ad6eef1184e00d1c1d0e3a41674d0a3688f9bbc707f86de9","abstract_canon_sha256":"34b2028fdd7d5484d0daef3de97018bba2af62509b6ce1a3e910216938e14baf"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:31:57.384342Z","signature_b64":"gZcfPr9uvlAGCrx8KEyB6ZC6VAi7J6MWWtHOmWDd3LOmezPADhWyfTWBFiOqP8Pf4UuM2A/S5aW5CmbS2oJACA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"93af8572524edd58455d3e475c148429f48e66f1dbcf24643274a67e6bbba4a4","last_reissued_at":"2026-07-05T10:31:57.383603Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:31:57.383603Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"TRUTH DECAY: Quantifying Multi-Turn Sycophancy in Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Aarav Jain, Aslihan Akalin, Joshua Liu, Kevin Zhu, Sean O'Brien, Soham Takuri, Srihan Vege, Vasu Sharma","submitted_at":"2025-02-04T06:56:32Z","abstract_excerpt":"Rapid improvements in large language models have unveiled a critical challenge in human-AI interaction: sycophancy. In this context, sycophancy refers to the tendency of models to excessively agree with or flatter users, often at the expense of factual accuracy. While previous studies have primarily analyzed this behavior in single-turn interactions, its persistence and evolution in multi-step conversations remain largely unexplored. We introduce TRUTH DECAY, a benchmark specifically designed to evaluate sycophancy in extended dialogues, where language models must navigate iterative user feedb"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.11656","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/2503.11656/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":"2503.11656","created_at":"2026-07-05T10:31:57.383694+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.11656v1","created_at":"2026-07-05T10:31:57.383694+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.11656","created_at":"2026-07-05T10:31:57.383694+00:00"},{"alias_kind":"pith_short_12","alias_value":"SOXYK4SSJ3OV","created_at":"2026-07-05T10:31:57.383694+00:00"},{"alias_kind":"pith_short_16","alias_value":"SOXYK4SSJ3OVQRK5","created_at":"2026-07-05T10:31:57.383694+00:00"},{"alias_kind":"pith_short_8","alias_value":"SOXYK4SS","created_at":"2026-07-05T10:31:57.383694+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.01071","citing_title":"MemSyco-Bench: Benchmarking Sycophancy in Agent Memory","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2607.01071","citing_title":"MemSyco-Bench: Benchmarking Sycophancy in Agent Memory","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2606.21296","citing_title":"Discriminatory Compliance: How LLMs Answer Queries from Protected Groups","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18799","citing_title":"ReCrit: Transition-Aware Reinforcement Learning for Scientific Critic Reasoning","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2510.07517","citing_title":"When Identity Skews Debate: Anonymization for Bias-Reduced Multi-Agent Reasoning","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05403","citing_title":"When Helpfulness Becomes Sycophancy: Sycophancy is a Boundary Failure Between Social Alignment and Epistemic Integrity in Large Language Models","ref_index":31,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/SOXYK4SSJ3OVQRK5HZDVYFEEFH","json":"https://pith.science/pith/SOXYK4SSJ3OVQRK5HZDVYFEEFH.json","graph_json":"https://pith.science/api/pith-number/SOXYK4SSJ3OVQRK5HZDVYFEEFH/graph.json","events_json":"https://pith.science/api/pith-number/SOXYK4SSJ3OVQRK5HZDVYFEEFH/events.json","paper":"https://pith.science/paper/SOXYK4SS"},"agent_actions":{"view_html":"https://pith.science/pith/SOXYK4SSJ3OVQRK5HZDVYFEEFH","download_json":"https://pith.science/pith/SOXYK4SSJ3OVQRK5HZDVYFEEFH.json","view_paper":"https://pith.science/paper/SOXYK4SS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.11656&json=true","fetch_graph":"https://pith.science/api/pith-number/SOXYK4SSJ3OVQRK5HZDVYFEEFH/graph.json","fetch_events":"https://pith.science/api/pith-number/SOXYK4SSJ3OVQRK5HZDVYFEEFH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SOXYK4SSJ3OVQRK5HZDVYFEEFH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SOXYK4SSJ3OVQRK5HZDVYFEEFH/action/storage_attestation","attest_author":"https://pith.science/pith/SOXYK4SSJ3OVQRK5HZDVYFEEFH/action/author_attestation","sign_citation":"https://pith.science/pith/SOXYK4SSJ3OVQRK5HZDVYFEEFH/action/citation_signature","submit_replication":"https://pith.science/pith/SOXYK4SSJ3OVQRK5HZDVYFEEFH/action/replication_record"}},"created_at":"2026-07-05T10:31:57.383694+00:00","updated_at":"2026-07-05T10:31:57.383694+00:00"}