{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:TPY4YTKNQ7RNBMX3EYXFZVKZGU","short_pith_number":"pith:TPY4YTKN","schema_version":"1.0","canonical_sha256":"9bf1cc4d4d87e2d0b2fb262e5cd559350b0785270ba61b32e6f397ebde51bce7","source":{"kind":"arxiv","id":"2406.11780","version":1},"attestation_state":"computed","paper":{"title":"Split, Unlearn, Merge: Leveraging Data Attributes for More Effective Unlearning in LLMs","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Dennis Wei, Farhan Ahmed, Inkit Padhi, Nathalie Baracaldo, Swanand Ravindra Kadhe","submitted_at":"2024-06-17T17:35:52Z","abstract_excerpt":"Large language models (LLMs) have shown to pose social and ethical risks such as generating toxic language or facilitating malicious use of hazardous knowledge. Machine unlearning is a promising approach to improve LLM safety by directly removing harmful behaviors and knowledge. In this paper, we propose \"SPlit, UNlearn, MerGE\" (SPUNGE), a framework that can be used with any unlearning method to amplify its effectiveness. SPUNGE leverages data attributes during unlearning by splitting unlearning data into subsets based on specific attribute values, unlearning each subset separately, and mergin"},"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":"2406.11780","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-17T17:35:52Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"fab9a24331a0949ecf938542ca673ba943b2255036f82a8127d3b5fff2fe6cbf","abstract_canon_sha256":"0c7f9a8c2d669db1df9995526d47ec115ae54911c8aa17bb18d5be4cde57b7b0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:32:57.899676Z","signature_b64":"liISodKV27ZkwvKgGRevIxLjZpjlULZVQDv8owzrX+Ab9LKNCgu5KYV6/WCO8aOC3RsRqOTw/ZhmT9YOakmgBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9bf1cc4d4d87e2d0b2fb262e5cd559350b0785270ba61b32e6f397ebde51bce7","last_reissued_at":"2026-07-05T08:32:57.899207Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:32:57.899207Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Split, Unlearn, Merge: Leveraging Data Attributes for More Effective Unlearning in LLMs","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Dennis Wei, Farhan Ahmed, Inkit Padhi, Nathalie Baracaldo, Swanand Ravindra Kadhe","submitted_at":"2024-06-17T17:35:52Z","abstract_excerpt":"Large language models (LLMs) have shown to pose social and ethical risks such as generating toxic language or facilitating malicious use of hazardous knowledge. Machine unlearning is a promising approach to improve LLM safety by directly removing harmful behaviors and knowledge. In this paper, we propose \"SPlit, UNlearn, MerGE\" (SPUNGE), a framework that can be used with any unlearning method to amplify its effectiveness. SPUNGE leverages data attributes during unlearning by splitting unlearning data into subsets based on specific attribute values, unlearning each subset separately, and mergin"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.11780","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/2406.11780/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":"2406.11780","created_at":"2026-07-05T08:32:57.899263+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.11780v1","created_at":"2026-07-05T08:32:57.899263+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.11780","created_at":"2026-07-05T08:32:57.899263+00:00"},{"alias_kind":"pith_short_12","alias_value":"TPY4YTKNQ7RN","created_at":"2026-07-05T08:32:57.899263+00:00"},{"alias_kind":"pith_short_16","alias_value":"TPY4YTKNQ7RNBMX3","created_at":"2026-07-05T08:32:57.899263+00:00"},{"alias_kind":"pith_short_8","alias_value":"TPY4YTKN","created_at":"2026-07-05T08:32:57.899263+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2506.20941","citing_title":"Revisiting the Past: Data Unlearning with Model State History","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17396","citing_title":"Representation-Guided Parameter-Efficient LLM Unlearning","ref_index":197,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TPY4YTKNQ7RNBMX3EYXFZVKZGU","json":"https://pith.science/pith/TPY4YTKNQ7RNBMX3EYXFZVKZGU.json","graph_json":"https://pith.science/api/pith-number/TPY4YTKNQ7RNBMX3EYXFZVKZGU/graph.json","events_json":"https://pith.science/api/pith-number/TPY4YTKNQ7RNBMX3EYXFZVKZGU/events.json","paper":"https://pith.science/paper/TPY4YTKN"},"agent_actions":{"view_html":"https://pith.science/pith/TPY4YTKNQ7RNBMX3EYXFZVKZGU","download_json":"https://pith.science/pith/TPY4YTKNQ7RNBMX3EYXFZVKZGU.json","view_paper":"https://pith.science/paper/TPY4YTKN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.11780&json=true","fetch_graph":"https://pith.science/api/pith-number/TPY4YTKNQ7RNBMX3EYXFZVKZGU/graph.json","fetch_events":"https://pith.science/api/pith-number/TPY4YTKNQ7RNBMX3EYXFZVKZGU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TPY4YTKNQ7RNBMX3EYXFZVKZGU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TPY4YTKNQ7RNBMX3EYXFZVKZGU/action/storage_attestation","attest_author":"https://pith.science/pith/TPY4YTKNQ7RNBMX3EYXFZVKZGU/action/author_attestation","sign_citation":"https://pith.science/pith/TPY4YTKNQ7RNBMX3EYXFZVKZGU/action/citation_signature","submit_replication":"https://pith.science/pith/TPY4YTKNQ7RNBMX3EYXFZVKZGU/action/replication_record"}},"created_at":"2026-07-05T08:32:57.899263+00:00","updated_at":"2026-07-05T08:32:57.899263+00:00"}