{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:FGJADUHTYVEONPI7DROLMFKQJN","short_pith_number":"pith:FGJADUHT","schema_version":"1.0","canonical_sha256":"299201d0f3c548e6bd1f1c5cb615504b6e80cbafd24f3175c730dd3627879324","source":{"kind":"arxiv","id":"2511.21214","version":4},"attestation_state":"computed","paper":{"title":"Self-Guided Adaptive Safety Alignment: Synthesizing and Internalizing Guidelines in Reasoning Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Dongyuan Lu, Jiaming Zhang, Jitao Sang, Yanxu Zhu, Yuhang Wang","submitted_at":"2025-11-26T09:44:32Z","abstract_excerpt":"Explicit safety policies can improve reasoning-model safety, but their effective coverage may lag behind evolving jailbreak strategies. We study whether a reasoning model can synthesize and internalize a task-specific safety guideline from a small set of harmful and benign examples. We introduce Self-Guided Adaptive Safety Alignment (SGASA). The model generates a guideline, refines it on its own errors, and selects a version by self-evaluation, which can then be applied in context or distilled into the model for guideline-free inference. Across two adversarial prompt datasets and three Qwen3 s"},"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":"2511.21214","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-11-26T09:44:32Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"bf389cd2ea25f7ad134087f0886e92caf3e750bc663f29c8783b8690def1ce56","abstract_canon_sha256":"6749bd0451675a75685ddd8b6d8470285e81d7ab6a965acb1987631e690e3aed"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-05T00:41:12.833809Z","signature_b64":"IoSxeeETNpVTUPXIznfwE4/bGfsyMM3baZXvCz57uY9OMhHHAkJkJkNSGJcTRJ0szM1JgG8rMleTz/BwqTiqBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"299201d0f3c548e6bd1f1c5cb615504b6e80cbafd24f3175c730dd3627879324","last_reissued_at":"2026-08-05T00:41:12.831192Z","signature_status":"signed_v1","first_computed_at":"2026-08-05T00:41:12.831192Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Self-Guided Adaptive Safety Alignment: Synthesizing and Internalizing Guidelines in Reasoning Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Dongyuan Lu, Jiaming Zhang, Jitao Sang, Yanxu Zhu, Yuhang Wang","submitted_at":"2025-11-26T09:44:32Z","abstract_excerpt":"Explicit safety policies can improve reasoning-model safety, but their effective coverage may lag behind evolving jailbreak strategies. We study whether a reasoning model can synthesize and internalize a task-specific safety guideline from a small set of harmful and benign examples. We introduce Self-Guided Adaptive Safety Alignment (SGASA). The model generates a guideline, refines it on its own errors, and selects a version by self-evaluation, which can then be applied in context or distilled into the model for guideline-free inference. Across two adversarial prompt datasets and three Qwen3 s"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2511.21214","kind":"arxiv","version":4},"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/2511.21214/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":"2511.21214","created_at":"2026-08-05T00:41:12.832064+00:00"},{"alias_kind":"arxiv_version","alias_value":"2511.21214v4","created_at":"2026-08-05T00:41:12.832064+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2511.21214","created_at":"2026-08-05T00:41:12.832064+00:00"},{"alias_kind":"pith_short_12","alias_value":"FGJADUHTYVEO","created_at":"2026-08-05T00:41:12.832064+00:00"},{"alias_kind":"pith_short_16","alias_value":"FGJADUHTYVEONPI7","created_at":"2026-08-05T00:41:12.832064+00:00"},{"alias_kind":"pith_short_8","alias_value":"FGJADUHT","created_at":"2026-08-05T00:41:12.832064+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/FGJADUHTYVEONPI7DROLMFKQJN","json":"https://pith.science/pith/FGJADUHTYVEONPI7DROLMFKQJN.json","graph_json":"https://pith.science/api/pith-number/FGJADUHTYVEONPI7DROLMFKQJN/graph.json","events_json":"https://pith.science/api/pith-number/FGJADUHTYVEONPI7DROLMFKQJN/events.json","paper":"https://pith.science/paper/FGJADUHT"},"agent_actions":{"view_html":"https://pith.science/pith/FGJADUHTYVEONPI7DROLMFKQJN","download_json":"https://pith.science/pith/FGJADUHTYVEONPI7DROLMFKQJN.json","view_paper":"https://pith.science/paper/FGJADUHT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2511.21214&json=true","fetch_graph":"https://pith.science/api/pith-number/FGJADUHTYVEONPI7DROLMFKQJN/graph.json","fetch_events":"https://pith.science/api/pith-number/FGJADUHTYVEONPI7DROLMFKQJN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FGJADUHTYVEONPI7DROLMFKQJN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FGJADUHTYVEONPI7DROLMFKQJN/action/storage_attestation","attest_author":"https://pith.science/pith/FGJADUHTYVEONPI7DROLMFKQJN/action/author_attestation","sign_citation":"https://pith.science/pith/FGJADUHTYVEONPI7DROLMFKQJN/action/citation_signature","submit_replication":"https://pith.science/pith/FGJADUHTYVEONPI7DROLMFKQJN/action/replication_record"}},"created_at":"2026-08-05T00:41:12.832064+00:00","updated_at":"2026-08-05T00:41:12.832064+00:00"}