{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:T6HXDXI5T6DCCNYEMDWBKLT24O","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":"c0ad3f258483854b11b7e8c6a5cfb15767e9be967efcd7e68a2d2e6a8ac57a69","cross_cats_sorted":["cs.AI","cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-04-16T15:46:02Z","title_canon_sha256":"9599830383a0f2641a0b445a59d5275738d96d4977a5d3214365f5153c26e668"},"schema_version":"1.0","source":{"id":"2604.15166","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.15166","created_at":"2026-05-20T02:05:43Z"},{"alias_kind":"arxiv_version","alias_value":"2604.15166v2","created_at":"2026-05-20T02:05:43Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.15166","created_at":"2026-05-20T02:05:43Z"},{"alias_kind":"pith_short_12","alias_value":"T6HXDXI5T6DC","created_at":"2026-05-20T02:05:43Z"},{"alias_kind":"pith_short_16","alias_value":"T6HXDXI5T6DCCNYE","created_at":"2026-05-20T02:05:43Z"},{"alias_kind":"pith_short_8","alias_value":"T6HXDXI5","created_at":"2026-05-20T02:05:43Z"}],"graph_snapshots":[{"event_id":"sha256:1646eb596f844654c1ebfbb2383e71cbc631926a765397f7644535b8b853460a","target":"graph","created_at":"2026-05-20T02:05:43Z","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":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Across MNIST, CIFAR-10, CIFAR-100, and Tiny ImageNet, and across convolutional and transformer architectures, DAMP more closely resembles the retraining gold standard than some of the prior methods, improving selective forgetting while better preserving retain-class performance and reducing residual forget-class structure in deep layers."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That forget directions extracted as residuals relative to retain-class prototypes at each layer accurately isolate the targeted knowledge, and that the parameter-free depth-aware scaling derived from probe separability will preserve utility without introducing new failure modes on retain classes."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"DAMP performs one-shot class unlearning by extracting and projecting out forget-specific residual directions at each network depth using class prototypes and a separability-derived scaling rule."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"By projecting out forget-specific directions layer by layer with depth-aware scaling, a closed-form method achieves class unlearning closer to full retraining than prior approaches."}],"snapshot_sha256":"b40c76d111c9bac4a0b4b8059e503eca22e0e488a746da4a466244dc9022aee2"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2604.15166/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Machine unlearning aims to remove targeted knowledge from a trained model without the cost of retraining from scratch. In class unlearning, however, reducing accuracy on forget classes does not necessarily imply true forgetting: forgotten information can remain encoded in internal representations, and apparent forgetting may arise from classifier-head suppression rather than representational removal. We show that existing class-unlearning methods often exhibit weak or negative selectivity, preserve forget-class structure in deep representations, or rely heavily on final-layer bias shifts. We t","authors_text":"Arman Hatami, Ilya E. Monosov, Romina Aalishah","cross_cats":["cs.AI","cs.LG"],"headline":"By projecting out forget-specific directions layer by layer with depth-aware scaling, a closed-form method achieves class unlearning closer to full retraining than prior approaches.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-04-16T15:46:02Z","title":"Class Unlearning via Depth-Aware Removal of Forget-Specific Directions"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.15166","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-10T11:42:48.804913Z","id":"ce474362-0a7f-4108-bf0c-ce88d88df969","model_set":{"reader":"grok-4.3"},"one_line_summary":"DAMP performs one-shot class unlearning by extracting and projecting out forget-specific residual directions at each network depth using class prototypes and a separability-derived scaling rule.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"By projecting out forget-specific directions layer by layer with depth-aware scaling, a closed-form method achieves class unlearning closer to full retraining than prior approaches.","strongest_claim":"Across MNIST, CIFAR-10, CIFAR-100, and Tiny ImageNet, and across convolutional and transformer architectures, DAMP more closely resembles the retraining gold standard than some of the prior methods, improving selective forgetting while better preserving retain-class performance and reducing residual forget-class structure in deep layers.","weakest_assumption":"That forget directions extracted as residuals relative to retain-class prototypes at each layer accurately isolate the targeted knowledge, and that the parameter-free depth-aware scaling derived from probe separability will preserve utility without introducing new failure modes on retain classes."}},"verdict_id":"ce474362-0a7f-4108-bf0c-ce88d88df969"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:bb7ef095050753823745f2fbf0fe0f9af4ede3bf63bd3fcab58248041748a389","target":"record","created_at":"2026-05-20T02:05:43Z","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":"c0ad3f258483854b11b7e8c6a5cfb15767e9be967efcd7e68a2d2e6a8ac57a69","cross_cats_sorted":["cs.AI","cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-04-16T15:46:02Z","title_canon_sha256":"9599830383a0f2641a0b445a59d5275738d96d4977a5d3214365f5153c26e668"},"schema_version":"1.0","source":{"id":"2604.15166","kind":"arxiv","version":2}},"canonical_sha256":"9f8f71dd1d9f8621370460ec152e7ae38ebf59870be78a60f56e61b6503c9373","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"9f8f71dd1d9f8621370460ec152e7ae38ebf59870be78a60f56e61b6503c9373","first_computed_at":"2026-05-20T02:05:43.203711Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T02:05:43.203711Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"hOoy4w14JwQN+29joNgmkVS+HyUbj/kgKF/5ITuVFLIqOwfYD/DbetMgs1SzUzF2vot0sWFf2dboFJNiTsTADg==","signature_status":"signed_v1","signed_at":"2026-05-20T02:05:43.204758Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.15166","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:bb7ef095050753823745f2fbf0fe0f9af4ede3bf63bd3fcab58248041748a389","sha256:1646eb596f844654c1ebfbb2383e71cbc631926a765397f7644535b8b853460a"],"state_sha256":"380f958e5e73ffb7a2592d5e58958ae22433210b106747b86ecd22986e533091"}