{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:QW74DFPW3I3RNNDQQK7ABCOL5H","short_pith_number":"pith:QW74DFPW","schema_version":"1.0","canonical_sha256":"85bfc195f6da3716b47082be0089cbe9e766ff0bf4649f1973ef0d0f911bc778","source":{"kind":"arxiv","id":"2506.08399","version":2},"attestation_state":"computed","paper":{"title":"SafeCoT: Improving VLM Safety with Minimal Reasoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Chaochao Lu, Chao Yang, Jiachen Ma, Zhanhui Zhou","submitted_at":"2025-06-10T03:13:50Z","abstract_excerpt":"Ensuring safe and appropriate responses from vision-language models (VLMs) remains a critical challenge, particularly in high-risk or ambiguous scenarios. We introduce SafeCoT, a lightweight, interpretable framework that leverages rule-based chain-of-thought (CoT) supervision to improve refusal behavior in VLMs. Unlike prior methods that rely on large-scale safety annotations or complex modeling, SafeCoT uses minimal supervision to help models reason about safety risks and make context-aware refusals. Experiments across multiple benchmarks show that SafeCoT significantly reduces overrefusal an"},"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":"2506.08399","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-06-10T03:13:50Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"078f42aa9d4e9c5fe5599c82e603d9777751c7aaace767406a572b0ada2c4efb","abstract_canon_sha256":"64a0628971119565267dc3548c546df368a1f2004f48d9df27855ea2dfed85d9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:19:29.458620Z","signature_b64":"Lk66uCynx7JsKl3bVMvcYylncPSoznQUbTd4AYqx4WtTO4ezQvOql/hyu/FwWnC3X3D5MTm550pmsXyS41YqDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"85bfc195f6da3716b47082be0089cbe9e766ff0bf4649f1973ef0d0f911bc778","last_reissued_at":"2026-07-05T11:19:29.457998Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:19:29.457998Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SafeCoT: Improving VLM Safety with Minimal Reasoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Chaochao Lu, Chao Yang, Jiachen Ma, Zhanhui Zhou","submitted_at":"2025-06-10T03:13:50Z","abstract_excerpt":"Ensuring safe and appropriate responses from vision-language models (VLMs) remains a critical challenge, particularly in high-risk or ambiguous scenarios. We introduce SafeCoT, a lightweight, interpretable framework that leverages rule-based chain-of-thought (CoT) supervision to improve refusal behavior in VLMs. Unlike prior methods that rely on large-scale safety annotations or complex modeling, SafeCoT uses minimal supervision to help models reason about safety risks and make context-aware refusals. Experiments across multiple benchmarks show that SafeCoT significantly reduces overrefusal an"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.08399","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/2506.08399/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":"2506.08399","created_at":"2026-07-05T11:19:29.458065+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.08399v2","created_at":"2026-07-05T11:19:29.458065+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.08399","created_at":"2026-07-05T11:19:29.458065+00:00"},{"alias_kind":"pith_short_12","alias_value":"QW74DFPW3I3R","created_at":"2026-07-05T11:19:29.458065+00:00"},{"alias_kind":"pith_short_16","alias_value":"QW74DFPW3I3RNNDQ","created_at":"2026-07-05T11:19:29.458065+00:00"},{"alias_kind":"pith_short_8","alias_value":"QW74DFPW","created_at":"2026-07-05T11:19:29.458065+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/QW74DFPW3I3RNNDQQK7ABCOL5H","json":"https://pith.science/pith/QW74DFPW3I3RNNDQQK7ABCOL5H.json","graph_json":"https://pith.science/api/pith-number/QW74DFPW3I3RNNDQQK7ABCOL5H/graph.json","events_json":"https://pith.science/api/pith-number/QW74DFPW3I3RNNDQQK7ABCOL5H/events.json","paper":"https://pith.science/paper/QW74DFPW"},"agent_actions":{"view_html":"https://pith.science/pith/QW74DFPW3I3RNNDQQK7ABCOL5H","download_json":"https://pith.science/pith/QW74DFPW3I3RNNDQQK7ABCOL5H.json","view_paper":"https://pith.science/paper/QW74DFPW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.08399&json=true","fetch_graph":"https://pith.science/api/pith-number/QW74DFPW3I3RNNDQQK7ABCOL5H/graph.json","fetch_events":"https://pith.science/api/pith-number/QW74DFPW3I3RNNDQQK7ABCOL5H/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QW74DFPW3I3RNNDQQK7ABCOL5H/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QW74DFPW3I3RNNDQQK7ABCOL5H/action/storage_attestation","attest_author":"https://pith.science/pith/QW74DFPW3I3RNNDQQK7ABCOL5H/action/author_attestation","sign_citation":"https://pith.science/pith/QW74DFPW3I3RNNDQQK7ABCOL5H/action/citation_signature","submit_replication":"https://pith.science/pith/QW74DFPW3I3RNNDQQK7ABCOL5H/action/replication_record"}},"created_at":"2026-07-05T11:19:29.458065+00:00","updated_at":"2026-07-05T11:19:29.458065+00:00"}