{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:QJ5G6AOYIYCALI5NNYBWLF2QRB","short_pith_number":"pith:QJ5G6AOY","schema_version":"1.0","canonical_sha256":"827a6f01d8460405a3ad6e0365975088777e5a0f5e713aa4e68cf63aa53d7555","source":{"kind":"arxiv","id":"2412.07097","version":1},"attestation_state":"computed","paper":{"title":"On Evaluating the Durability of Safeguards for Open-Weight LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Boyi Wei, Luxi He, Matthew Jagielski, Milad Nasr, Nicholas Carlini, Peter Henderson, Prateek Mittal, Tinghao Xie, Xiangyu Qi, Yangsibo Huang","submitted_at":"2024-12-10T01:30:32Z","abstract_excerpt":"Stakeholders -- from model developers to policymakers -- seek to minimize the dual-use risks of large language models (LLMs). An open challenge to this goal is whether technical safeguards can impede the misuse of LLMs, even when models are customizable via fine-tuning or when model weights are fully open. In response, several recent studies have proposed methods to produce durable LLM safeguards for open-weight LLMs that can withstand adversarial modifications of the model's weights via fine-tuning. This holds the promise of raising adversaries' costs even under strong threat models where adv"},"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":"2412.07097","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2024-12-10T01:30:32Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"e8611ea233b83230901da795a21219a6f13a0674194aaeea5b083e9a3bed5539","abstract_canon_sha256":"8cec4f3854658bc079022f825abd60c62cc5198d3db392d4445d0caef7c598be"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:47:03.540382Z","signature_b64":"lCoS2TzZtczMEDfUdj2owHWX40Y1SkbSOtMEEiRTDPdYlfEyCMygg7DPk+gpbNRQccbx++ZgJIIrurznxGzjAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"827a6f01d8460405a3ad6e0365975088777e5a0f5e713aa4e68cf63aa53d7555","last_reissued_at":"2026-07-05T09:47:03.539898Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:47:03.539898Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"On Evaluating the Durability of Safeguards for Open-Weight LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Boyi Wei, Luxi He, Matthew Jagielski, Milad Nasr, Nicholas Carlini, Peter Henderson, Prateek Mittal, Tinghao Xie, Xiangyu Qi, Yangsibo Huang","submitted_at":"2024-12-10T01:30:32Z","abstract_excerpt":"Stakeholders -- from model developers to policymakers -- seek to minimize the dual-use risks of large language models (LLMs). An open challenge to this goal is whether technical safeguards can impede the misuse of LLMs, even when models are customizable via fine-tuning or when model weights are fully open. In response, several recent studies have proposed methods to produce durable LLM safeguards for open-weight LLMs that can withstand adversarial modifications of the model's weights via fine-tuning. This holds the promise of raising adversaries' costs even under strong threat models where adv"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.07097","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/2412.07097/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":"2412.07097","created_at":"2026-07-05T09:47:03.539950+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.07097v1","created_at":"2026-07-05T09:47:03.539950+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.07097","created_at":"2026-07-05T09:47:03.539950+00:00"},{"alias_kind":"pith_short_12","alias_value":"QJ5G6AOYIYCA","created_at":"2026-07-05T09:47:03.539950+00:00"},{"alias_kind":"pith_short_16","alias_value":"QJ5G6AOYIYCALI5N","created_at":"2026-07-05T09:47:03.539950+00:00"},{"alias_kind":"pith_short_8","alias_value":"QJ5G6AOY","created_at":"2026-07-05T09:47:03.539950+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.25073","citing_title":"Security in the Fine-Tuning Lifecycle of Large Language Models: Threats, Defenses,Evaluation, and Future Directions","ref_index":57,"is_internal_anchor":false},{"citing_arxiv_id":"2605.25073","citing_title":"Security in the Fine-Tuning Lifecycle of Large Language Models: Threats, Defenses,Evaluation, and Future Directions","ref_index":58,"is_internal_anchor":false},{"citing_arxiv_id":"2506.06414","citing_title":"Benchmarking Misuse Mitigation Against Covert Adversaries","ref_index":65,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QJ5G6AOYIYCALI5NNYBWLF2QRB","json":"https://pith.science/pith/QJ5G6AOYIYCALI5NNYBWLF2QRB.json","graph_json":"https://pith.science/api/pith-number/QJ5G6AOYIYCALI5NNYBWLF2QRB/graph.json","events_json":"https://pith.science/api/pith-number/QJ5G6AOYIYCALI5NNYBWLF2QRB/events.json","paper":"https://pith.science/paper/QJ5G6AOY"},"agent_actions":{"view_html":"https://pith.science/pith/QJ5G6AOYIYCALI5NNYBWLF2QRB","download_json":"https://pith.science/pith/QJ5G6AOYIYCALI5NNYBWLF2QRB.json","view_paper":"https://pith.science/paper/QJ5G6AOY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.07097&json=true","fetch_graph":"https://pith.science/api/pith-number/QJ5G6AOYIYCALI5NNYBWLF2QRB/graph.json","fetch_events":"https://pith.science/api/pith-number/QJ5G6AOYIYCALI5NNYBWLF2QRB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QJ5G6AOYIYCALI5NNYBWLF2QRB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QJ5G6AOYIYCALI5NNYBWLF2QRB/action/storage_attestation","attest_author":"https://pith.science/pith/QJ5G6AOYIYCALI5NNYBWLF2QRB/action/author_attestation","sign_citation":"https://pith.science/pith/QJ5G6AOYIYCALI5NNYBWLF2QRB/action/citation_signature","submit_replication":"https://pith.science/pith/QJ5G6AOYIYCALI5NNYBWLF2QRB/action/replication_record"}},"created_at":"2026-07-05T09:47:03.539950+00:00","updated_at":"2026-07-05T09:47:03.539950+00:00"}