{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:7YF2LCIPNXOL736HX2I2SGIHS5","short_pith_number":"pith:7YF2LCIP","schema_version":"1.0","canonical_sha256":"fe0ba5890f6ddcbfefc7be91a919079749564a04b5cb5656038f665865a9888a","source":{"kind":"arxiv","id":"2603.26846","version":2},"attestation_state":"computed","paper":{"title":"Stable Reasoning, Unstable Responses: Mitigating LLM Deception via Stability Asymmetry","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Guoxi Zhang, Jiawei Chen, Jingwei Yi, Juntao Dai, Lang Qin, Tianzhuo Yang, Yaodong Yang","submitted_at":"2026-03-27T09:47:57Z","abstract_excerpt":"As Large Language Models (LLMs) expand in capability and application scope, their trustworthiness becomes critical. A vital risk is intrinsic deception, wherein models strategically mislead users to achieve their own objectives. Existing alignment approaches based on chain-of-thought (CoT) monitoring supervise explicit reasoning traces. However, under optimization pressure, models are incentivized to conceal deceptive reasoning, rendering semantic supervision fundamentally unreliable. Grounded in cognitive psychology, we hypothesize that a deceptive LLM maintains a stable internal belief in it"},"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":"2603.26846","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-03-27T09:47:57Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"55bff1aba6de3959a83cae2af3605a67105ee8b4f3a02dfebc5a8c4e8778c721","abstract_canon_sha256":"7e2dede3fe07a47c7679bb3ad8f0b25fcb86e9a39dc8ee3afc68ae07137aff63"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-08T01:04:02.604848Z","signature_b64":"d0BRIl1uOEKLSSXi6TnXwqP5rEo1mrccerH5XlAppYqgRSWsaDHJ96nXJCa27yfrsHkjbXj7lBdf9FNYEBPOCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fe0ba5890f6ddcbfefc7be91a919079749564a04b5cb5656038f665865a9888a","last_reissued_at":"2026-06-08T01:04:02.603847Z","signature_status":"signed_v1","first_computed_at":"2026-06-08T01:04:02.603847Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Stable Reasoning, Unstable Responses: Mitigating LLM Deception via Stability Asymmetry","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Guoxi Zhang, Jiawei Chen, Jingwei Yi, Juntao Dai, Lang Qin, Tianzhuo Yang, Yaodong Yang","submitted_at":"2026-03-27T09:47:57Z","abstract_excerpt":"As Large Language Models (LLMs) expand in capability and application scope, their trustworthiness becomes critical. A vital risk is intrinsic deception, wherein models strategically mislead users to achieve their own objectives. Existing alignment approaches based on chain-of-thought (CoT) monitoring supervise explicit reasoning traces. However, under optimization pressure, models are incentivized to conceal deceptive reasoning, rendering semantic supervision fundamentally unreliable. Grounded in cognitive psychology, we hypothesize that a deceptive LLM maintains a stable internal belief in it"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.26846","kind":"arxiv","version":2},"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/2603.26846/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":"2603.26846","created_at":"2026-06-08T01:04:02.603997+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.26846v2","created_at":"2026-06-08T01:04:02.603997+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.26846","created_at":"2026-06-08T01:04:02.603997+00:00"},{"alias_kind":"pith_short_12","alias_value":"7YF2LCIPNXOL","created_at":"2026-06-08T01:04:02.603997+00:00"},{"alias_kind":"pith_short_16","alias_value":"7YF2LCIPNXOL736H","created_at":"2026-06-08T01:04:02.603997+00:00"},{"alias_kind":"pith_short_8","alias_value":"7YF2LCIP","created_at":"2026-06-08T01:04:02.603997+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7YF2LCIPNXOL736HX2I2SGIHS5","json":"https://pith.science/pith/7YF2LCIPNXOL736HX2I2SGIHS5.json","graph_json":"https://pith.science/api/pith-number/7YF2LCIPNXOL736HX2I2SGIHS5/graph.json","events_json":"https://pith.science/api/pith-number/7YF2LCIPNXOL736HX2I2SGIHS5/events.json","paper":"https://pith.science/paper/7YF2LCIP"},"agent_actions":{"view_html":"https://pith.science/pith/7YF2LCIPNXOL736HX2I2SGIHS5","download_json":"https://pith.science/pith/7YF2LCIPNXOL736HX2I2SGIHS5.json","view_paper":"https://pith.science/paper/7YF2LCIP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.26846&json=true","fetch_graph":"https://pith.science/api/pith-number/7YF2LCIPNXOL736HX2I2SGIHS5/graph.json","fetch_events":"https://pith.science/api/pith-number/7YF2LCIPNXOL736HX2I2SGIHS5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7YF2LCIPNXOL736HX2I2SGIHS5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7YF2LCIPNXOL736HX2I2SGIHS5/action/storage_attestation","attest_author":"https://pith.science/pith/7YF2LCIPNXOL736HX2I2SGIHS5/action/author_attestation","sign_citation":"https://pith.science/pith/7YF2LCIPNXOL736HX2I2SGIHS5/action/citation_signature","submit_replication":"https://pith.science/pith/7YF2LCIPNXOL736HX2I2SGIHS5/action/replication_record"}},"created_at":"2026-06-08T01:04:02.603997+00:00","updated_at":"2026-06-08T01:04:02.603997+00:00"}