{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:N7W2C7JY2AVQN4U55LFAFIYYDP","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":"cc38e5c2edd8b9b9cb1a78d797b37ca59e850ecca91f64d5ba0bfd8e1de4f4f6","cross_cats_sorted":["cs.AI","cs.CL","cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2024-06-28T17:05:46Z","title_canon_sha256":"fce39fc5ff1df767363bbd4f437064d693f0e71c1bae755a2bf3b4777e045f80"},"schema_version":"1.0","source":{"id":"2406.20053","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2406.20053","created_at":"2026-07-05T08:37:58Z"},{"alias_kind":"arxiv_version","alias_value":"2406.20053v1","created_at":"2026-07-05T08:37:58Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.20053","created_at":"2026-07-05T08:37:58Z"},{"alias_kind":"pith_short_12","alias_value":"N7W2C7JY2AVQ","created_at":"2026-07-05T08:37:58Z"},{"alias_kind":"pith_short_16","alias_value":"N7W2C7JY2AVQN4U5","created_at":"2026-07-05T08:37:58Z"},{"alias_kind":"pith_short_8","alias_value":"N7W2C7JY","created_at":"2026-07-05T08:37:58Z"}],"graph_snapshots":[{"event_id":"sha256:36a7236d430e7c45bbac0a7495bf25e1a20eac3a3c4e49016e61cfa548d1737f","target":"graph","created_at":"2026-07-05T08:37:58Z","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/2406.20053/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Black-box finetuning is an emerging interface for adapting state-of-the-art language models to user needs. However, such access may also let malicious actors undermine model safety. To demonstrate the challenge of defending finetuning interfaces, we introduce covert malicious finetuning, a method to compromise model safety via finetuning while evading detection. Our method constructs a malicious dataset where every individual datapoint appears innocuous, but finetuning on the dataset teaches the model to respond to encoded harmful requests with encoded harmful responses. Applied to GPT-4, our ","authors_text":"Alexander Wei, Danny Halawi, Eric Wallace, Jacob Steinhardt, Nika Haghtalab, Tony T. Wang","cross_cats":["cs.AI","cs.CL","cs.LG"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2024-06-28T17:05:46Z","title":"Covert Malicious Finetuning: Challenges in Safeguarding LLM Adaptation"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.20053","kind":"arxiv","version":1},"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":"sha256:905ce9c0cf79365e496fa97ace269899b884bf685c6e6c3bdcd2b6976b2f8c9a","target":"record","created_at":"2026-07-05T08:37:58Z","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":"cc38e5c2edd8b9b9cb1a78d797b37ca59e850ecca91f64d5ba0bfd8e1de4f4f6","cross_cats_sorted":["cs.AI","cs.CL","cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2024-06-28T17:05:46Z","title_canon_sha256":"fce39fc5ff1df767363bbd4f437064d693f0e71c1bae755a2bf3b4777e045f80"},"schema_version":"1.0","source":{"id":"2406.20053","kind":"arxiv","version":1}},"canonical_sha256":"6feda17d38d02b06f29deaca02a3181bd907c330a0a3ad193a771ef97d041f23","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"6feda17d38d02b06f29deaca02a3181bd907c330a0a3ad193a771ef97d041f23","first_computed_at":"2026-07-05T08:37:58.511200Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T08:37:58.511200Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"vrcWOsyRt9JK5EmETrd+/VZW6vxMUJJklOWD7iJdGwkAsNeQk2ue4Tr/Lj+x7NDILaHHmVZvCNZVSJ6NcYDkDg==","signature_status":"signed_v1","signed_at":"2026-07-05T08:37:58.511688Z","signed_message":"canonical_sha256_bytes"},"source_id":"2406.20053","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:905ce9c0cf79365e496fa97ace269899b884bf685c6e6c3bdcd2b6976b2f8c9a","sha256:36a7236d430e7c45bbac0a7495bf25e1a20eac3a3c4e49016e61cfa548d1737f"],"state_sha256":"30a8272b1b9194c5d62f2d851ad5cf3e38450fbb8ce58ff0c9db516b9ecc54f4"}