{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:2XXBQPRUIS2QI7SXUBYAPU3NT5","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"d41c0b506c361ad90ceb84580c874b6e582e8f994d0ff9e67a37722cb87fc837","cross_cats_sorted":["cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-05-23T13:51:55Z","title_canon_sha256":"b80411dddc6468e80a7a0a898f8d49761747df37fdb1295b55b23baf3dcb81cf"},"schema_version":"1.0","source":{"id":"2405.14577","kind":"arxiv","version":4}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2405.14577","created_at":"2026-07-05T09:28:49Z"},{"alias_kind":"arxiv_version","alias_value":"2405.14577v4","created_at":"2026-07-05T09:28:49Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.14577","created_at":"2026-07-05T09:28:49Z"},{"alias_kind":"pith_short_12","alias_value":"2XXBQPRUIS2Q","created_at":"2026-07-05T09:28:49Z"},{"alias_kind":"pith_short_16","alias_value":"2XXBQPRUIS2QI7SX","created_at":"2026-07-05T09:28:49Z"},{"alias_kind":"pith_short_8","alias_value":"2XXBQPRU","created_at":"2026-07-05T09:28:49Z"}],"graph_snapshots":[{"event_id":"sha256:505f241fbe26a2efa4390ae67c8db5a59d199945c34b9ece960b1674a85fb6b2","target":"graph","created_at":"2026-07-05T09:28:49Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2405.14577/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Releasing open-source large language models (LLMs) presents a dual-use risk since bad actors can easily fine-tune these models for harmful purposes. Even without the open release of weights, weight stealing and fine-tuning APIs make closed models vulnerable to harmful fine-tuning attacks (HFAs). While safety measures like preventing jailbreaks and improving safety guardrails are important, such measures can easily be reversed through fine-tuning. In this work, we propose Representation Noising (RepNoise), a defence mechanism that operates even when attackers have access to the weights. RepNois","authors_text":"Carsten Maple, David Atanasov, Domenic Rosati, Frank Rudzicz, Hassan Sajjad, Jan Wehner, Kai Williams, {\\L}ukasz Bartoszcze, Robie Gonzales, Subhabrata Majumdar","cross_cats":["cs.LG"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-05-23T13:51:55Z","title":"Representation Noising: A Defence Mechanism Against Harmful Finetuning"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.14577","kind":"arxiv","version":4},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:dc4f3688f2cd8b6617d8bc95f6782d6ec273a09f93c0a60ea19989b17c627e8b","target":"record","created_at":"2026-07-05T09:28:49Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"d41c0b506c361ad90ceb84580c874b6e582e8f994d0ff9e67a37722cb87fc837","cross_cats_sorted":["cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-05-23T13:51:55Z","title_canon_sha256":"b80411dddc6468e80a7a0a898f8d49761747df37fdb1295b55b23baf3dcb81cf"},"schema_version":"1.0","source":{"id":"2405.14577","kind":"arxiv","version":4}},"canonical_sha256":"d5ee183e3444b5047e57a07007d36d9f587af13cbfd946c71725705bd1c43248","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"d5ee183e3444b5047e57a07007d36d9f587af13cbfd946c71725705bd1c43248","first_computed_at":"2026-07-05T09:28:49.966104Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T09:28:49.966104Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"vN21S2r61opE8sqZxk2QzhUsQakYbZGbc+5xprOAs4FuXpu/2blhA+b/jsPZejqQNxDmuwGn6cI5RLKTSeG7AA==","signature_status":"signed_v1","signed_at":"2026-07-05T09:28:49.966605Z","signed_message":"canonical_sha256_bytes"},"source_id":"2405.14577","source_kind":"arxiv","source_version":4}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:dc4f3688f2cd8b6617d8bc95f6782d6ec273a09f93c0a60ea19989b17c627e8b","sha256:505f241fbe26a2efa4390ae67c8db5a59d199945c34b9ece960b1674a85fb6b2"],"state_sha256":"412d1a2251bea41ebe013a28ccc1db86cfc39859c69521f579b5e95c98b0e339"}