{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:BQJN2YQA5JGBBCJLO5F67GA7MF","short_pith_number":"pith:BQJN2YQA","schema_version":"1.0","canonical_sha256":"0c12dd6200ea4c10892b774bef981f616df410a6ed090936d0d31b583c7a00ee","source":{"kind":"arxiv","id":"2508.13743","version":1},"attestation_state":"computed","paper":{"title":"Sycophancy under Pressure: Evaluating and Mitigating Sycophantic Bias via Adversarial Dialogues in Scientific QA","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chunyi Li, Dandan Zhu, Guangtao Zhai, Kaiwei Zhang, Qi Jia, Wei Sun, Xiangyang Zhu, Zijian Chen","submitted_at":"2025-08-19T11:30:52Z","abstract_excerpt":"Large language models (LLMs), while increasingly used in domains requiring factual rigor, often display a troubling behavior: sycophancy, the tendency to align with user beliefs regardless of correctness. This tendency is reinforced by preference-based alignment techniques that optimize for user satisfaction but can undermine truthfulness. While relatively benign in casual dialogue, sycophancy poses serious risks in high-stakes settings such as scientific question answering (QA), where model outputs may shape collaborative reasoning, decision-making, and knowledge formation. Despite its import"},"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":"2508.13743","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-08-19T11:30:52Z","cross_cats_sorted":[],"title_canon_sha256":"ec5f7e84dc990cfd80b2a553679f5e9efa8b7d11aa6542b5d0474525bb675246","abstract_canon_sha256":"5220015cccc7262568215819ecd08e218f35a9f13f7807a969ef57b35e90049f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:56:06.997341Z","signature_b64":"1PFrGCHA7CY9wpLmIh/NU2VdBX4GXGcOCLWtdcj84/UB+QPR9h0hr57tDHZaJ437e7PhMl7M9/pzdE4AP0T3Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0c12dd6200ea4c10892b774bef981f616df410a6ed090936d0d31b583c7a00ee","last_reissued_at":"2026-07-05T11:56:06.996803Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:56:06.996803Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Sycophancy under Pressure: Evaluating and Mitigating Sycophantic Bias via Adversarial Dialogues in Scientific QA","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chunyi Li, Dandan Zhu, Guangtao Zhai, Kaiwei Zhang, Qi Jia, Wei Sun, Xiangyang Zhu, Zijian Chen","submitted_at":"2025-08-19T11:30:52Z","abstract_excerpt":"Large language models (LLMs), while increasingly used in domains requiring factual rigor, often display a troubling behavior: sycophancy, the tendency to align with user beliefs regardless of correctness. This tendency is reinforced by preference-based alignment techniques that optimize for user satisfaction but can undermine truthfulness. While relatively benign in casual dialogue, sycophancy poses serious risks in high-stakes settings such as scientific question answering (QA), where model outputs may shape collaborative reasoning, decision-making, and knowledge formation. Despite its import"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.13743","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/2508.13743/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":"2508.13743","created_at":"2026-07-05T11:56:06.996866+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.13743v1","created_at":"2026-07-05T11:56:06.996866+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.13743","created_at":"2026-07-05T11:56:06.996866+00:00"},{"alias_kind":"pith_short_12","alias_value":"BQJN2YQA5JGB","created_at":"2026-07-05T11:56:06.996866+00:00"},{"alias_kind":"pith_short_16","alias_value":"BQJN2YQA5JGBBCJL","created_at":"2026-07-05T11:56:06.996866+00:00"},{"alias_kind":"pith_short_8","alias_value":"BQJN2YQA","created_at":"2026-07-05T11:56:06.996866+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.23932","citing_title":"When Correct Beliefs Collapse: Epistemic Resilience of LLMs under Clinical Pressure","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2606.10949","citing_title":"Recalling Too Well: Sycophancy Evaluation and Mitigation in Memory-Augmented Models","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06306","citing_title":"Decomposing Factual Sycophancy in Language Models: How Size and Instruction Tuning Shape Robustness","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2510.07517","citing_title":"When Identity Skews Debate: Anonymization for Bias-Reduced Multi-Agent Reasoning","ref_index":51,"is_internal_anchor":false},{"citing_arxiv_id":"2604.02423","citing_title":"SWAY: A Counterfactual Computational Linguistic Approach to Measuring and Mitigating Sycophancy","ref_index":30,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BQJN2YQA5JGBBCJLO5F67GA7MF","json":"https://pith.science/pith/BQJN2YQA5JGBBCJLO5F67GA7MF.json","graph_json":"https://pith.science/api/pith-number/BQJN2YQA5JGBBCJLO5F67GA7MF/graph.json","events_json":"https://pith.science/api/pith-number/BQJN2YQA5JGBBCJLO5F67GA7MF/events.json","paper":"https://pith.science/paper/BQJN2YQA"},"agent_actions":{"view_html":"https://pith.science/pith/BQJN2YQA5JGBBCJLO5F67GA7MF","download_json":"https://pith.science/pith/BQJN2YQA5JGBBCJLO5F67GA7MF.json","view_paper":"https://pith.science/paper/BQJN2YQA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.13743&json=true","fetch_graph":"https://pith.science/api/pith-number/BQJN2YQA5JGBBCJLO5F67GA7MF/graph.json","fetch_events":"https://pith.science/api/pith-number/BQJN2YQA5JGBBCJLO5F67GA7MF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BQJN2YQA5JGBBCJLO5F67GA7MF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BQJN2YQA5JGBBCJLO5F67GA7MF/action/storage_attestation","attest_author":"https://pith.science/pith/BQJN2YQA5JGBBCJLO5F67GA7MF/action/author_attestation","sign_citation":"https://pith.science/pith/BQJN2YQA5JGBBCJLO5F67GA7MF/action/citation_signature","submit_replication":"https://pith.science/pith/BQJN2YQA5JGBBCJLO5F67GA7MF/action/replication_record"}},"created_at":"2026-07-05T11:56:06.996866+00:00","updated_at":"2026-07-05T11:56:06.996866+00:00"}