{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:UBVH3FUIFZPIWTF25HBL7QVNTL","short_pith_number":"pith:UBVH3FUI","canonical_record":{"source":{"id":"2410.22086","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-10-29T14:41:44Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"49278e9acff5b7d5782476c65f6cfe93a48f960e29eefbdeb2110186f8d966e2","abstract_canon_sha256":"c06f4e1aa5372e3c6949e5f79927ed55815ca71a14f0d2efe6d2c553e07f13d7"},"schema_version":"1.0"},"canonical_sha256":"a06a7d96882e5e8b4cbae9c2bfc2ad9aebc4b1bcb7ece0d90eb761c38836098c","source":{"kind":"arxiv","id":"2410.22086","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2410.22086","created_at":"2026-07-05T10:58:46Z"},{"alias_kind":"arxiv_version","alias_value":"2410.22086v3","created_at":"2026-07-05T10:58:46Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.22086","created_at":"2026-07-05T10:58:46Z"},{"alias_kind":"pith_short_12","alias_value":"UBVH3FUIFZPI","created_at":"2026-07-05T10:58:46Z"},{"alias_kind":"pith_short_16","alias_value":"UBVH3FUIFZPIWTF2","created_at":"2026-07-05T10:58:46Z"},{"alias_kind":"pith_short_8","alias_value":"UBVH3FUI","created_at":"2026-07-05T10:58:46Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:UBVH3FUIFZPIWTF25HBL7QVNTL","target":"record","payload":{"canonical_record":{"source":{"id":"2410.22086","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-10-29T14:41:44Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"49278e9acff5b7d5782476c65f6cfe93a48f960e29eefbdeb2110186f8d966e2","abstract_canon_sha256":"c06f4e1aa5372e3c6949e5f79927ed55815ca71a14f0d2efe6d2c553e07f13d7"},"schema_version":"1.0"},"canonical_sha256":"a06a7d96882e5e8b4cbae9c2bfc2ad9aebc4b1bcb7ece0d90eb761c38836098c","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:58:46.261594Z","signature_b64":"B1Sh9rwe76/3SkgqJxqrHjxWYCpIKTL8Lg32H102dUF8xuka1Hwwp3yTxrq1y/4/5fx2u3lf+7ZoYnNQDRVuAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a06a7d96882e5e8b4cbae9c2bfc2ad9aebc4b1bcb7ece0d90eb761c38836098c","last_reissued_at":"2026-07-05T10:58:46.261116Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:58:46.261116Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2410.22086","source_version":3,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:58:46Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"AwJTFGkd5FY73SptScpk9EDRACEkcjEL9EzTDZVoUQVu41c4jaeMO3m4uPODsrs4CAshEbwxYsRbOjkItfcVCg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T19:47:04.281430Z"},"content_sha256":"30874650f5f6478a9f1cc22f80ab8737f634656126305629e1fcf358f2818ab8","schema_version":"1.0","event_id":"sha256:30874650f5f6478a9f1cc22f80ab8737f634656126305629e1fcf358f2818ab8"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:UBVH3FUIFZPIWTF25HBL7QVNTL","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Unlearning as multi-task optimization: A normalized gradient difference approach with an adaptive learning rate","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Anil Ramakrishna, Bhanukiran Vinzamuri, Kai-Wei Chang, Mingyi Hong, Volkan Cevher, Xiaomeng Jin, Zhiqi Bu","submitted_at":"2024-10-29T14:41:44Z","abstract_excerpt":"Machine unlearning has been used to remove unwanted knowledge acquired by large language models (LLMs). In this paper, we examine machine unlearning from an optimization perspective, framing it as a regularized multi-task optimization problem, where one task optimizes a forgetting objective and another optimizes the model performance. In particular, we introduce a normalized gradient difference (NGDiff) algorithm, enabling us to have better control over the trade-off between the objectives, while integrating a new, automatic learning rate scheduler. We provide a theoretical analysis and empiri"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.22086","kind":"arxiv","version":3},"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/2410.22086/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:58:46Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"xJ5t4Y/O72R/PHmSd3FgyyiHgYGDe82h34A5NONxOdF3pNdnezkNCHA3FAkqFkQKNp4558yc5siEdxTzJl5LBA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T19:47:04.282163Z"},"content_sha256":"1fd4e1f27b14ee2b10f00fe6617332f317b76cc342c8bc898e1a82911f6667c3","schema_version":"1.0","event_id":"sha256:1fd4e1f27b14ee2b10f00fe6617332f317b76cc342c8bc898e1a82911f6667c3"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/UBVH3FUIFZPIWTF25HBL7QVNTL/bundle.json","state_url":"https://pith.science/pith/UBVH3FUIFZPIWTF25HBL7QVNTL/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/UBVH3FUIFZPIWTF25HBL7QVNTL/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-05T19:47:04Z","links":{"resolver":"https://pith.science/pith/UBVH3FUIFZPIWTF25HBL7QVNTL","bundle":"https://pith.science/pith/UBVH3FUIFZPIWTF25HBL7QVNTL/bundle.json","state":"https://pith.science/pith/UBVH3FUIFZPIWTF25HBL7QVNTL/state.json","well_known_bundle":"https://pith.science/.well-known/pith/UBVH3FUIFZPIWTF25HBL7QVNTL/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:UBVH3FUIFZPIWTF25HBL7QVNTL","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":"c06f4e1aa5372e3c6949e5f79927ed55815ca71a14f0d2efe6d2c553e07f13d7","cross_cats_sorted":["cs.CL"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-10-29T14:41:44Z","title_canon_sha256":"49278e9acff5b7d5782476c65f6cfe93a48f960e29eefbdeb2110186f8d966e2"},"schema_version":"1.0","source":{"id":"2410.22086","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2410.22086","created_at":"2026-07-05T10:58:46Z"},{"alias_kind":"arxiv_version","alias_value":"2410.22086v3","created_at":"2026-07-05T10:58:46Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.22086","created_at":"2026-07-05T10:58:46Z"},{"alias_kind":"pith_short_12","alias_value":"UBVH3FUIFZPI","created_at":"2026-07-05T10:58:46Z"},{"alias_kind":"pith_short_16","alias_value":"UBVH3FUIFZPIWTF2","created_at":"2026-07-05T10:58:46Z"},{"alias_kind":"pith_short_8","alias_value":"UBVH3FUI","created_at":"2026-07-05T10:58:46Z"}],"graph_snapshots":[{"event_id":"sha256:1fd4e1f27b14ee2b10f00fe6617332f317b76cc342c8bc898e1a82911f6667c3","target":"graph","created_at":"2026-07-05T10:58:46Z","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/2410.22086/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Machine unlearning has been used to remove unwanted knowledge acquired by large language models (LLMs). In this paper, we examine machine unlearning from an optimization perspective, framing it as a regularized multi-task optimization problem, where one task optimizes a forgetting objective and another optimizes the model performance. In particular, we introduce a normalized gradient difference (NGDiff) algorithm, enabling us to have better control over the trade-off between the objectives, while integrating a new, automatic learning rate scheduler. We provide a theoretical analysis and empiri","authors_text":"Anil Ramakrishna, Bhanukiran Vinzamuri, Kai-Wei Chang, Mingyi Hong, Volkan Cevher, Xiaomeng Jin, Zhiqi Bu","cross_cats":["cs.CL"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-10-29T14:41:44Z","title":"Unlearning as multi-task optimization: A normalized gradient difference approach with an adaptive learning rate"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.22086","kind":"arxiv","version":3},"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":"sha256:30874650f5f6478a9f1cc22f80ab8737f634656126305629e1fcf358f2818ab8","target":"record","created_at":"2026-07-05T10:58:46Z","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":"c06f4e1aa5372e3c6949e5f79927ed55815ca71a14f0d2efe6d2c553e07f13d7","cross_cats_sorted":["cs.CL"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-10-29T14:41:44Z","title_canon_sha256":"49278e9acff5b7d5782476c65f6cfe93a48f960e29eefbdeb2110186f8d966e2"},"schema_version":"1.0","source":{"id":"2410.22086","kind":"arxiv","version":3}},"canonical_sha256":"a06a7d96882e5e8b4cbae9c2bfc2ad9aebc4b1bcb7ece0d90eb761c38836098c","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"a06a7d96882e5e8b4cbae9c2bfc2ad9aebc4b1bcb7ece0d90eb761c38836098c","first_computed_at":"2026-07-05T10:58:46.261116Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:58:46.261116Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"B1Sh9rwe76/3SkgqJxqrHjxWYCpIKTL8Lg32H102dUF8xuka1Hwwp3yTxrq1y/4/5fx2u3lf+7ZoYnNQDRVuAg==","signature_status":"signed_v1","signed_at":"2026-07-05T10:58:46.261594Z","signed_message":"canonical_sha256_bytes"},"source_id":"2410.22086","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:30874650f5f6478a9f1cc22f80ab8737f634656126305629e1fcf358f2818ab8","sha256:1fd4e1f27b14ee2b10f00fe6617332f317b76cc342c8bc898e1a82911f6667c3"],"state_sha256":"fc6eaeac4a04c7192d0ab24104b30792a4bc46f45b0a073ed457cd4e323c3896"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"8vgwJR9xjyv2tl9M5Wtt3U0tlhaZ819fkcrgggbIk3OOi6tURzCjw5zd1IK003ul+u8Qb13/K3q/hCKRFIWyDQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-05T19:47:04.289563Z","bundle_sha256":"c22bcd4f09cc7582709179bddf96d6b09e987cd465f35546cb9a99b00fd83c5e"}}