{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:76XMYXCVO7LF5GHBRYBDYFUWZR","short_pith_number":"pith:76XMYXCV","schema_version":"1.0","canonical_sha256":"ffaecc5c5577d65e98e18e023c1696cc65af1adfdcb289fcce3ebba9850ab8cc","source":{"kind":"arxiv","id":"2406.12670","version":2},"attestation_state":"computed","paper":{"title":"Stealth edits to large language models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Alexander Bastounis, Alexander N. Gorban, Desmond J. Higham, Ivan Y. Tyukin, Oliver J. Sutton, Qinghua Zhou, Wei Wang","submitted_at":"2024-06-18T14:43:18Z","abstract_excerpt":"We reveal the theoretical foundations of techniques for editing large language models, and present new methods which can do so without requiring retraining. Our theoretical insights show that a single metric (a measure of the intrinsic dimension of the model's features) can be used to assess a model's editability and reveals its previously unrecognised susceptibility to malicious stealth attacks. This metric is fundamental to predicting the success of a variety of editing approaches, and reveals new bridges between disparate families of editing methods. We collectively refer to these as stealt"},"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":"2406.12670","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-06-18T14:43:18Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"6c538ca48e93136e42b774bfdad1ece714303df0f448d9b7bd3d0e75972f38c4","abstract_canon_sha256":"acc7c2cfdffb683f75a3a235c9d1d4dd45f4fc93812ead846022817f929036de"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:28:32.490476Z","signature_b64":"dCroH06Z4H9jdzzDssNqiPRE/mMpukDrJBZ9lphIssP+YA1nPCsmOE09uarLlIw8iU87OAKYa0h+bpsFBpPCDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ffaecc5c5577d65e98e18e023c1696cc65af1adfdcb289fcce3ebba9850ab8cc","last_reissued_at":"2026-07-05T09:28:32.489959Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:28:32.489959Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Stealth edits to large language models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Alexander Bastounis, Alexander N. Gorban, Desmond J. Higham, Ivan Y. Tyukin, Oliver J. Sutton, Qinghua Zhou, Wei Wang","submitted_at":"2024-06-18T14:43:18Z","abstract_excerpt":"We reveal the theoretical foundations of techniques for editing large language models, and present new methods which can do so without requiring retraining. Our theoretical insights show that a single metric (a measure of the intrinsic dimension of the model's features) can be used to assess a model's editability and reveals its previously unrecognised susceptibility to malicious stealth attacks. This metric is fundamental to predicting the success of a variety of editing approaches, and reveals new bridges between disparate families of editing methods. We collectively refer to these as stealt"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.12670","kind":"arxiv","version":2},"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/2406.12670/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":"2406.12670","created_at":"2026-07-05T09:28:32.490011+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.12670v2","created_at":"2026-07-05T09:28:32.490011+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.12670","created_at":"2026-07-05T09:28:32.490011+00:00"},{"alias_kind":"pith_short_12","alias_value":"76XMYXCVO7LF","created_at":"2026-07-05T09:28:32.490011+00:00"},{"alias_kind":"pith_short_16","alias_value":"76XMYXCVO7LF5GHB","created_at":"2026-07-05T09:28:32.490011+00:00"},{"alias_kind":"pith_short_8","alias_value":"76XMYXCV","created_at":"2026-07-05T09:28:32.490011+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.14242","citing_title":"Implicit Reasoning Steering via Concept Chaining","ref_index":162,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/76XMYXCVO7LF5GHBRYBDYFUWZR","json":"https://pith.science/pith/76XMYXCVO7LF5GHBRYBDYFUWZR.json","graph_json":"https://pith.science/api/pith-number/76XMYXCVO7LF5GHBRYBDYFUWZR/graph.json","events_json":"https://pith.science/api/pith-number/76XMYXCVO7LF5GHBRYBDYFUWZR/events.json","paper":"https://pith.science/paper/76XMYXCV"},"agent_actions":{"view_html":"https://pith.science/pith/76XMYXCVO7LF5GHBRYBDYFUWZR","download_json":"https://pith.science/pith/76XMYXCVO7LF5GHBRYBDYFUWZR.json","view_paper":"https://pith.science/paper/76XMYXCV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.12670&json=true","fetch_graph":"https://pith.science/api/pith-number/76XMYXCVO7LF5GHBRYBDYFUWZR/graph.json","fetch_events":"https://pith.science/api/pith-number/76XMYXCVO7LF5GHBRYBDYFUWZR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/76XMYXCVO7LF5GHBRYBDYFUWZR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/76XMYXCVO7LF5GHBRYBDYFUWZR/action/storage_attestation","attest_author":"https://pith.science/pith/76XMYXCVO7LF5GHBRYBDYFUWZR/action/author_attestation","sign_citation":"https://pith.science/pith/76XMYXCVO7LF5GHBRYBDYFUWZR/action/citation_signature","submit_replication":"https://pith.science/pith/76XMYXCVO7LF5GHBRYBDYFUWZR/action/replication_record"}},"created_at":"2026-07-05T09:28:32.490011+00:00","updated_at":"2026-07-05T09:28:32.490011+00:00"}