{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:7HDC4DQJ7YXAKR2KCNQMEXOJRK","short_pith_number":"pith:7HDC4DQJ","schema_version":"1.0","canonical_sha256":"f9c62e0e09fe2e05474a1360c25dc98a80d738b7bafdb8b3b9165cb0866fa70f","source":{"kind":"arxiv","id":"2310.02664","version":2},"attestation_state":"computed","paper":{"title":"On Memorization in Diffusion Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Chao Du, Chongxuan Li, Min Lin, Tianyu Pang, Xiangming Gu, Ye Wang","submitted_at":"2023-10-04T09:04:20Z","abstract_excerpt":"Due to their capacity to generate novel and high-quality samples, diffusion models have attracted significant research interest in recent years. Notably, the typical training objective of diffusion models, i.e., denoising score matching, has a closed-form optimal solution that can only generate training data replicating samples. This indicates that a memorization behavior is theoretically expected, which contradicts the common generalization ability of state-of-the-art diffusion models, and thus calls for a deeper understanding. Looking into this, we first observe that memorization behaviors t"},"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":"2310.02664","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-10-04T09:04:20Z","cross_cats_sorted":["cs.AI","cs.CV"],"title_canon_sha256":"1d9b676f842d5b62397ca898f8d4c0008e37fd4ed116c0158ae743189b0ee70d","abstract_canon_sha256":"aa597235258c0caa20c88d75dd3db80e78192774e0b2a31e0ace71858338fb19"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:17:05.366123Z","signature_b64":"CvVL05C4L+qMOrVZyaY/I7kW+Fai9fafF72baZQ0hg+MQ2Hf+EviBkMX+33+MLCkvVgQzJfZnO5lTLAhbhLgAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f9c62e0e09fe2e05474a1360c25dc98a80d738b7bafdb8b3b9165cb0866fa70f","last_reissued_at":"2026-07-05T10:17:05.365615Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:17:05.365615Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"On Memorization in Diffusion Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Chao Du, Chongxuan Li, Min Lin, Tianyu Pang, Xiangming Gu, Ye Wang","submitted_at":"2023-10-04T09:04:20Z","abstract_excerpt":"Due to their capacity to generate novel and high-quality samples, diffusion models have attracted significant research interest in recent years. Notably, the typical training objective of diffusion models, i.e., denoising score matching, has a closed-form optimal solution that can only generate training data replicating samples. This indicates that a memorization behavior is theoretically expected, which contradicts the common generalization ability of state-of-the-art diffusion models, and thus calls for a deeper understanding. Looking into this, we first observe that memorization behaviors t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.02664","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/2310.02664/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":"2310.02664","created_at":"2026-07-05T10:17:05.365682+00:00"},{"alias_kind":"arxiv_version","alias_value":"2310.02664v2","created_at":"2026-07-05T10:17:05.365682+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.02664","created_at":"2026-07-05T10:17:05.365682+00:00"},{"alias_kind":"pith_short_12","alias_value":"7HDC4DQJ7YXA","created_at":"2026-07-05T10:17:05.365682+00:00"},{"alias_kind":"pith_short_16","alias_value":"7HDC4DQJ7YXAKR2K","created_at":"2026-07-05T10:17:05.365682+00:00"},{"alias_kind":"pith_short_8","alias_value":"7HDC4DQJ","created_at":"2026-07-05T10:17:05.365682+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":14,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.08041","citing_title":"An exact information theory of generalization phase transitions in Bayesian diffusion models","ref_index":16,"is_internal_anchor":true},{"citing_arxiv_id":"2606.20155","citing_title":"NAMESAKES: Probing Identity Memorization in Text-to-Image Models","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2606.08554","citing_title":"A Theoretical Analysis of Memory and Overfitting Phenomena in Stochastic Interpolation Models","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06624","citing_title":"Principles and Practice of Deep Representation Learning: or a Mathematical Theory of Memory","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31991","citing_title":"Amplifying Membership Signal Through Chained Regeneration","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2508.00756","citing_title":"LeakyCLIP: Extracting Training Data from CLIP","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2603.13419","citing_title":"Diffusion Models Memorize in Training -- and Generalize in Inference","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17500","citing_title":"The Silent Brush: Evaluating Artistic Style Leakage in AI Art Generation","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2506.13763","citing_title":"Diagnosing and Improving Diffusion Models by Estimating the Optimal Loss Value","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2509.25003","citing_title":"Score-based Membership Inference on Diffusion Models","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2511.05940","citing_title":"A PDE Perspective on Generative Diffusion Models","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10439","citing_title":"Filtering Memorization from Parameter-Space in Diffusion Models","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06367","citing_title":"The Interplay of Data Structure and Imbalance in the Learning Dynamics of Diffusion Models","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21860","citing_title":"Transient Turn Injection: Exposing Stateless Multi-Turn Vulnerabilities in Large Language Models","ref_index":24,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7HDC4DQJ7YXAKR2KCNQMEXOJRK","json":"https://pith.science/pith/7HDC4DQJ7YXAKR2KCNQMEXOJRK.json","graph_json":"https://pith.science/api/pith-number/7HDC4DQJ7YXAKR2KCNQMEXOJRK/graph.json","events_json":"https://pith.science/api/pith-number/7HDC4DQJ7YXAKR2KCNQMEXOJRK/events.json","paper":"https://pith.science/paper/7HDC4DQJ"},"agent_actions":{"view_html":"https://pith.science/pith/7HDC4DQJ7YXAKR2KCNQMEXOJRK","download_json":"https://pith.science/pith/7HDC4DQJ7YXAKR2KCNQMEXOJRK.json","view_paper":"https://pith.science/paper/7HDC4DQJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2310.02664&json=true","fetch_graph":"https://pith.science/api/pith-number/7HDC4DQJ7YXAKR2KCNQMEXOJRK/graph.json","fetch_events":"https://pith.science/api/pith-number/7HDC4DQJ7YXAKR2KCNQMEXOJRK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7HDC4DQJ7YXAKR2KCNQMEXOJRK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7HDC4DQJ7YXAKR2KCNQMEXOJRK/action/storage_attestation","attest_author":"https://pith.science/pith/7HDC4DQJ7YXAKR2KCNQMEXOJRK/action/author_attestation","sign_citation":"https://pith.science/pith/7HDC4DQJ7YXAKR2KCNQMEXOJRK/action/citation_signature","submit_replication":"https://pith.science/pith/7HDC4DQJ7YXAKR2KCNQMEXOJRK/action/replication_record"}},"created_at":"2026-07-05T10:17:05.365682+00:00","updated_at":"2026-07-05T10:17:05.365682+00:00"}