{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:OPGGFYYDJVAXIBWXSYQG2TZBXT","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":"fae7690b6445f3880277fa506720f0791fe49d30d89b6973ad86d906b3bd7913","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-10-04T13:25:32Z","title_canon_sha256":"48c29298b0e71f2653c460efa777a02204db5183e3e7373d6d7177399e30c6ec"},"schema_version":"1.0","source":{"id":"2410.03415","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2410.03415","created_at":"2026-07-05T10:23:23Z"},{"alias_kind":"arxiv_version","alias_value":"2410.03415v2","created_at":"2026-07-05T10:23:23Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.03415","created_at":"2026-07-05T10:23:23Z"},{"alias_kind":"pith_short_12","alias_value":"OPGGFYYDJVAX","created_at":"2026-07-05T10:23:23Z"},{"alias_kind":"pith_short_16","alias_value":"OPGGFYYDJVAXIBWX","created_at":"2026-07-05T10:23:23Z"},{"alias_kind":"pith_short_8","alias_value":"OPGGFYYD","created_at":"2026-07-05T10:23:23Z"}],"graph_snapshots":[{"event_id":"sha256:71ce694fa1a6835cf37658836ea66b5f8299104b634ff1e5e60c38b3b0887d1d","target":"graph","created_at":"2026-07-05T10:23:23Z","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/2410.03415/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Training a language model to be both helpful and harmless requires careful calibration of refusal behaviours: Models should refuse to follow malicious instructions or give harmful advice (e.g.\"how do I kill someone?\"), but they should not refuse safe requests, even if they superficially resemble unsafe ones (e.g. \"how do I kill a Python process?\"). Avoiding such false refusal, as prior work has shown, is challenging even for highly-capable language models. In this paper, we propose a simple and surgical method for mitigating false refusal in language models via single vector ablation. For a gi","authors_text":"Barbara Plank, Chengzhi Hu, Paul R\\\"ottger, Xinpeng Wang","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-10-04T13:25:32Z","title":"Surgical, Cheap, and Flexible: Mitigating False Refusal in Language Models via Single Vector Ablation"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.03415","kind":"arxiv","version":2},"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":"sha256:959ac1ee52a740d4c2d9d6967e7f4d5f5b9f160b7c24c4250cff024daf64bdef","target":"record","created_at":"2026-07-05T10:23:23Z","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":"fae7690b6445f3880277fa506720f0791fe49d30d89b6973ad86d906b3bd7913","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-10-04T13:25:32Z","title_canon_sha256":"48c29298b0e71f2653c460efa777a02204db5183e3e7373d6d7177399e30c6ec"},"schema_version":"1.0","source":{"id":"2410.03415","kind":"arxiv","version":2}},"canonical_sha256":"73cc62e3034d417406d796206d4f21bcc33ce8679455a13448e27df3db4bcf90","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"73cc62e3034d417406d796206d4f21bcc33ce8679455a13448e27df3db4bcf90","first_computed_at":"2026-07-05T10:23:23.592409Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:23:23.592409Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"F07A/UJbRkXDVj34zN9Wkk5u8u9WO/qnpa9ViZZCdBAnknIM25U0PwX+do1PhbmjoqJkoP/z3tAINtALYQTRCw==","signature_status":"signed_v1","signed_at":"2026-07-05T10:23:23.593102Z","signed_message":"canonical_sha256_bytes"},"source_id":"2410.03415","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:959ac1ee52a740d4c2d9d6967e7f4d5f5b9f160b7c24c4250cff024daf64bdef","sha256:71ce694fa1a6835cf37658836ea66b5f8299104b634ff1e5e60c38b3b0887d1d"],"state_sha256":"32de20b0cea3772c046e2c324e1ede33349e9f9e41ad848ba707ebae028bbb74"}