{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:UZLH2CCUOXDNRJJOWPL6BM6TCY","short_pith_number":"pith:UZLH2CCU","schema_version":"1.0","canonical_sha256":"a6567d085475c6d8a52eb3d7e0b3d3160599c77f07748e2d39e3aaa4422e9bfc","source":{"kind":"arxiv","id":"2401.05779","version":4},"attestation_state":"computed","paper":{"title":"Erasing Undesirable Influence in Diffusion Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Jing Wu, Mehrtash Harandi, Munawar Hayat, Trung Le","submitted_at":"2024-01-11T09:30:36Z","abstract_excerpt":"Diffusion models are highly effective at generating high-quality images but pose risks, such as the unintentional generation of NSFW (not safe for work) content. Although various techniques have been proposed to mitigate unwanted influences in diffusion models while preserving overall performance, achieving a balance between these goals remains challenging. In this work, we introduce EraseDiff, an algorithm designed to preserve the utility of the diffusion model on retained data while removing the unwanted information associated with the data to be forgotten. Our approach formulates this task "},"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":"2401.05779","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-01-11T09:30:36Z","cross_cats_sorted":[],"title_canon_sha256":"35aa0566c408edd7b7071fff12a86de5b2a134bca5187f352c3cca35d4d093c1","abstract_canon_sha256":"b0ae1342f5460b4c67381e9464189bb20f6942c09427eea9c520b72e56f17e49"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:37:51.477256Z","signature_b64":"5Cl+MLv/Hp0649vcwJjwj4twIEMioqacD8tyXt+GXje7zEDSGPIIvoTWNE49XfO/8d2OstkmOqVN2ORPZ38RBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a6567d085475c6d8a52eb3d7e0b3d3160599c77f07748e2d39e3aaa4422e9bfc","last_reissued_at":"2026-07-05T09:37:51.476756Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:37:51.476756Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Erasing Undesirable Influence in Diffusion Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Jing Wu, Mehrtash Harandi, Munawar Hayat, Trung Le","submitted_at":"2024-01-11T09:30:36Z","abstract_excerpt":"Diffusion models are highly effective at generating high-quality images but pose risks, such as the unintentional generation of NSFW (not safe for work) content. Although various techniques have been proposed to mitigate unwanted influences in diffusion models while preserving overall performance, achieving a balance between these goals remains challenging. In this work, we introduce EraseDiff, an algorithm designed to preserve the utility of the diffusion model on retained data while removing the unwanted information associated with the data to be forgotten. Our approach formulates this task "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2401.05779","kind":"arxiv","version":4},"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/2401.05779/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":"2401.05779","created_at":"2026-07-05T09:37:51.476818+00:00"},{"alias_kind":"arxiv_version","alias_value":"2401.05779v4","created_at":"2026-07-05T09:37:51.476818+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2401.05779","created_at":"2026-07-05T09:37:51.476818+00:00"},{"alias_kind":"pith_short_12","alias_value":"UZLH2CCUOXDN","created_at":"2026-07-05T09:37:51.476818+00:00"},{"alias_kind":"pith_short_16","alias_value":"UZLH2CCUOXDNRJJO","created_at":"2026-07-05T09:37:51.476818+00:00"},{"alias_kind":"pith_short_8","alias_value":"UZLH2CCU","created_at":"2026-07-05T09:37:51.476818+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07907","citing_title":"Multimodal Unlearning Across Vision, Language, Video, and Audio: Survey of Methods, Datasets, and Benchmarks","ref_index":84,"is_internal_anchor":true},{"citing_arxiv_id":"2606.24192","citing_title":"Co-occurring associated retained concepts in Diffusion Unlearning","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2605.26332","citing_title":"Erased but Exploitable: Black-box Embedding-Aware Prompting Against Unlearned Text-to-Image Diffusion Models","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25119","citing_title":"Evaluation without Generation: Non-Generative Assessment of Harmful Model Specialization with Applications to CSAM","ref_index":63,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04575","citing_title":"Erasure or Erosion? Evaluating Compositional Degradation in Unlearned Text-To-Image Diffusion Models","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2604.15829","citing_title":"Beyond Text Prompts: Precise Concept Erasure through Text-Image Collaboration","ref_index":43,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UZLH2CCUOXDNRJJOWPL6BM6TCY","json":"https://pith.science/pith/UZLH2CCUOXDNRJJOWPL6BM6TCY.json","graph_json":"https://pith.science/api/pith-number/UZLH2CCUOXDNRJJOWPL6BM6TCY/graph.json","events_json":"https://pith.science/api/pith-number/UZLH2CCUOXDNRJJOWPL6BM6TCY/events.json","paper":"https://pith.science/paper/UZLH2CCU"},"agent_actions":{"view_html":"https://pith.science/pith/UZLH2CCUOXDNRJJOWPL6BM6TCY","download_json":"https://pith.science/pith/UZLH2CCUOXDNRJJOWPL6BM6TCY.json","view_paper":"https://pith.science/paper/UZLH2CCU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2401.05779&json=true","fetch_graph":"https://pith.science/api/pith-number/UZLH2CCUOXDNRJJOWPL6BM6TCY/graph.json","fetch_events":"https://pith.science/api/pith-number/UZLH2CCUOXDNRJJOWPL6BM6TCY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UZLH2CCUOXDNRJJOWPL6BM6TCY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UZLH2CCUOXDNRJJOWPL6BM6TCY/action/storage_attestation","attest_author":"https://pith.science/pith/UZLH2CCUOXDNRJJOWPL6BM6TCY/action/author_attestation","sign_citation":"https://pith.science/pith/UZLH2CCUOXDNRJJOWPL6BM6TCY/action/citation_signature","submit_replication":"https://pith.science/pith/UZLH2CCUOXDNRJJOWPL6BM6TCY/action/replication_record"}},"created_at":"2026-07-05T09:37:51.476818+00:00","updated_at":"2026-07-05T09:37:51.476818+00:00"}