{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:HSYCO3STBT55JANEUL2ZVUETNZ","short_pith_number":"pith:HSYCO3ST","schema_version":"1.0","canonical_sha256":"3cb0276e530cfbd481a4a2f59ad0936e5a866f1231e955c4e26021c7d7f07a7a","source":{"kind":"arxiv","id":"2602.13562","version":2},"attestation_state":"computed","paper":{"title":"Mitigating the Safety-utility Trade-off in LLM Alignment via Adaptive Safe Context Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.CR","authors_text":"Jian Liang, Lu Wang, Minzheng Wang, Ran He, Yanbo Wang, Yongcan Yu","submitted_at":"2026-02-14T02:37:36Z","abstract_excerpt":"While reasoning models have achieved remarkable success in complex reasoning tasks, their increasing power necessitates stringent safety measures. For safety alignment, the core challenge lies in the inherent trade-off between safety and utility. However, prevailing alignment strategies typically construct CoT training data with explicit safety rules via context distillation. This approach inadvertently limits reasoning capabilities by creating a rigid association between rule memorization and refusal. To mitigate the safety-utility trade-off, we propose the Adaptive Safe Context Learning~(ASC"},"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":"2602.13562","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2026-02-14T02:37:36Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"86660332751e7b92e4f291d70360c363e0a6ceda08076ee28dd634983738c606","abstract_canon_sha256":"607faf1c53fe89a326133a9a9ee8c9b3fd528a5ab3f7abd372a47a8f54c18085"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-30T02:17:16.108985Z","signature_b64":"4n90bCL6lP9bBPywj7PRMMErFif1lh62h6KltixNHlsh9Z7XKnNKm3l15rYdmqaY7YPPfe2y1xZfDHnPgl7cDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3cb0276e530cfbd481a4a2f59ad0936e5a866f1231e955c4e26021c7d7f07a7a","last_reissued_at":"2026-06-30T02:17:16.108425Z","signature_status":"signed_v1","first_computed_at":"2026-06-30T02:17:16.108425Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mitigating the Safety-utility Trade-off in LLM Alignment via Adaptive Safe Context Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.CR","authors_text":"Jian Liang, Lu Wang, Minzheng Wang, Ran He, Yanbo Wang, Yongcan Yu","submitted_at":"2026-02-14T02:37:36Z","abstract_excerpt":"While reasoning models have achieved remarkable success in complex reasoning tasks, their increasing power necessitates stringent safety measures. For safety alignment, the core challenge lies in the inherent trade-off between safety and utility. However, prevailing alignment strategies typically construct CoT training data with explicit safety rules via context distillation. This approach inadvertently limits reasoning capabilities by creating a rigid association between rule memorization and refusal. To mitigate the safety-utility trade-off, we propose the Adaptive Safe Context Learning~(ASC"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.13562","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/2602.13562/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":"2602.13562","created_at":"2026-06-30T02:17:16.108507+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.13562v2","created_at":"2026-06-30T02:17:16.108507+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.13562","created_at":"2026-06-30T02:17:16.108507+00:00"},{"alias_kind":"pith_short_12","alias_value":"HSYCO3STBT55","created_at":"2026-06-30T02:17:16.108507+00:00"},{"alias_kind":"pith_short_16","alias_value":"HSYCO3STBT55JANE","created_at":"2026-06-30T02:17:16.108507+00:00"},{"alias_kind":"pith_short_8","alias_value":"HSYCO3ST","created_at":"2026-06-30T02:17:16.108507+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2604.14142","citing_title":"From $P(y|x)$ to $P(y)$: Investigating Reinforcement Learning in Pre-train Space","ref_index":60,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HSYCO3STBT55JANEUL2ZVUETNZ","json":"https://pith.science/pith/HSYCO3STBT55JANEUL2ZVUETNZ.json","graph_json":"https://pith.science/api/pith-number/HSYCO3STBT55JANEUL2ZVUETNZ/graph.json","events_json":"https://pith.science/api/pith-number/HSYCO3STBT55JANEUL2ZVUETNZ/events.json","paper":"https://pith.science/paper/HSYCO3ST"},"agent_actions":{"view_html":"https://pith.science/pith/HSYCO3STBT55JANEUL2ZVUETNZ","download_json":"https://pith.science/pith/HSYCO3STBT55JANEUL2ZVUETNZ.json","view_paper":"https://pith.science/paper/HSYCO3ST","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.13562&json=true","fetch_graph":"https://pith.science/api/pith-number/HSYCO3STBT55JANEUL2ZVUETNZ/graph.json","fetch_events":"https://pith.science/api/pith-number/HSYCO3STBT55JANEUL2ZVUETNZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HSYCO3STBT55JANEUL2ZVUETNZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HSYCO3STBT55JANEUL2ZVUETNZ/action/storage_attestation","attest_author":"https://pith.science/pith/HSYCO3STBT55JANEUL2ZVUETNZ/action/author_attestation","sign_citation":"https://pith.science/pith/HSYCO3STBT55JANEUL2ZVUETNZ/action/citation_signature","submit_replication":"https://pith.science/pith/HSYCO3STBT55JANEUL2ZVUETNZ/action/replication_record"}},"created_at":"2026-06-30T02:17:16.108507+00:00","updated_at":"2026-06-30T02:17:16.108507+00:00"}