{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:IIXE4FOU7DUZRGIYUIWLBKCTVZ","short_pith_number":"pith:IIXE4FOU","schema_version":"1.0","canonical_sha256":"422e4e15d4f8e9989918a22cb0a853ae58b48f6ae7b4cde75673c76f388640df","source":{"kind":"arxiv","id":"2410.10014","version":1},"attestation_state":"computed","paper":{"title":"Safety-Aware Fine-Tuning of Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Hyeong Kyu Choi, Xuefeng Du, Yixuan Li","submitted_at":"2024-10-13T21:24:25Z","abstract_excerpt":"Fine-tuning Large Language Models (LLMs) has emerged as a common practice for tailoring models to individual needs and preferences. The choice of datasets for fine-tuning can be diverse, introducing safety concerns regarding the potential inclusion of harmful data samples. Manually filtering or avoiding such samples, however, can be labor-intensive and subjective. To address these difficulties, we propose a novel Safety-Aware Fine-Tuning (SAFT) framework designed to automatically detect and remove potentially harmful data, by leveraging a scoring function that exploits the subspace information"},"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":"2410.10014","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-10-13T21:24:25Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"33dcbad41928f423b10af9640fccc5cec7e2d302cca26bc02efa52bc1312886d","abstract_canon_sha256":"3229a73bf65c72beba95dfdc361481446a9a7d69b3226777d612f84e9ff2194e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:20:09.473481Z","signature_b64":"9BUZ++/DfEUkzd9Lx80rwN4OuQQYSTz/9hPSa4O9ee2hQwSp9KB1ujwoQMO1uDmyQUU0HlmOp7gvUxw0iGZwBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"422e4e15d4f8e9989918a22cb0a853ae58b48f6ae7b4cde75673c76f388640df","last_reissued_at":"2026-07-05T09:20:09.473041Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:20:09.473041Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Safety-Aware Fine-Tuning of Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Hyeong Kyu Choi, Xuefeng Du, Yixuan Li","submitted_at":"2024-10-13T21:24:25Z","abstract_excerpt":"Fine-tuning Large Language Models (LLMs) has emerged as a common practice for tailoring models to individual needs and preferences. The choice of datasets for fine-tuning can be diverse, introducing safety concerns regarding the potential inclusion of harmful data samples. Manually filtering or avoiding such samples, however, can be labor-intensive and subjective. To address these difficulties, we propose a novel Safety-Aware Fine-Tuning (SAFT) framework designed to automatically detect and remove potentially harmful data, by leveraging a scoring function that exploits the subspace information"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.10014","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/2410.10014/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":"2410.10014","created_at":"2026-07-05T09:20:09.473097+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.10014v1","created_at":"2026-07-05T09:20:09.473097+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.10014","created_at":"2026-07-05T09:20:09.473097+00:00"},{"alias_kind":"pith_short_12","alias_value":"IIXE4FOU7DUZ","created_at":"2026-07-05T09:20:09.473097+00:00"},{"alias_kind":"pith_short_16","alias_value":"IIXE4FOU7DUZRGIY","created_at":"2026-07-05T09:20:09.473097+00:00"},{"alias_kind":"pith_short_8","alias_value":"IIXE4FOU","created_at":"2026-07-05T09:20:09.473097+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.01738","citing_title":"THRD: A Training-Free Multi-Turn Defense Framework for Jailbreak Attacks on Large Language Models","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2409.18169","citing_title":"Harmful Fine-tuning Attacks and Defenses for Large Language Models: A Survey","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14194","citing_title":"GradShield: Alignment Preserving Finetuning","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18946","citing_title":"Reasoning Structure Matters for Safety Alignment of Reasoning Models","ref_index":52,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IIXE4FOU7DUZRGIYUIWLBKCTVZ","json":"https://pith.science/pith/IIXE4FOU7DUZRGIYUIWLBKCTVZ.json","graph_json":"https://pith.science/api/pith-number/IIXE4FOU7DUZRGIYUIWLBKCTVZ/graph.json","events_json":"https://pith.science/api/pith-number/IIXE4FOU7DUZRGIYUIWLBKCTVZ/events.json","paper":"https://pith.science/paper/IIXE4FOU"},"agent_actions":{"view_html":"https://pith.science/pith/IIXE4FOU7DUZRGIYUIWLBKCTVZ","download_json":"https://pith.science/pith/IIXE4FOU7DUZRGIYUIWLBKCTVZ.json","view_paper":"https://pith.science/paper/IIXE4FOU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.10014&json=true","fetch_graph":"https://pith.science/api/pith-number/IIXE4FOU7DUZRGIYUIWLBKCTVZ/graph.json","fetch_events":"https://pith.science/api/pith-number/IIXE4FOU7DUZRGIYUIWLBKCTVZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IIXE4FOU7DUZRGIYUIWLBKCTVZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IIXE4FOU7DUZRGIYUIWLBKCTVZ/action/storage_attestation","attest_author":"https://pith.science/pith/IIXE4FOU7DUZRGIYUIWLBKCTVZ/action/author_attestation","sign_citation":"https://pith.science/pith/IIXE4FOU7DUZRGIYUIWLBKCTVZ/action/citation_signature","submit_replication":"https://pith.science/pith/IIXE4FOU7DUZRGIYUIWLBKCTVZ/action/replication_record"}},"created_at":"2026-07-05T09:20:09.473097+00:00","updated_at":"2026-07-05T09:20:09.473097+00:00"}