{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:4CFSWDKON4D6UJCEQ44TA567N3","short_pith_number":"pith:4CFSWDKO","schema_version":"1.0","canonical_sha256":"e08b2b0d4e6f07ea244487393077df6ef56166600ba971acd1c88158bbd45eda","source":{"kind":"arxiv","id":"2505.16959","version":2},"attestation_state":"computed","paper":{"title":"Bigger Isn't Always Memorizing: Early Stopping Overparameterized Diffusion Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Alessandro Favero, Antonio Sclocchi, Matthieu Wyart","submitted_at":"2025-05-22T17:40:08Z","abstract_excerpt":"Diffusion probabilistic models have become a cornerstone of modern generative AI, yet the mechanisms underlying their generalization remain poorly understood. In fact, if these models were perfectly minimizing their training loss, they would just generate data belonging to their training set, i.e., memorize, as empirically found in the overparameterized regime. We revisit this view by showing that, in highly overparameterized diffusion models, generalization in natural data domains is progressively achieved during training before the onset of memorization. Our results, ranging from image to la"},"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":"2505.16959","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-22T17:40:08Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"cfaea2fd5abcd9aab4ceb5a0f6a8110d6cb1a69866ff83e284442e6b6c4ee621","abstract_canon_sha256":"ac07738ce3f9f9cec5730014a0156f770cd0a241ede316433d0a384044039351"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:02:46.428494Z","signature_b64":"GNj8Wkn7p6c823iyrtdVBl0bjTdLT8KbzqPVefP6xMQ9CjPtXq95EziJLrWISiFfYbUiOpYTHPeJQc0aBBrlDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e08b2b0d4e6f07ea244487393077df6ef56166600ba971acd1c88158bbd45eda","last_reissued_at":"2026-07-05T12:02:46.427920Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:02:46.427920Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Bigger Isn't Always Memorizing: Early Stopping Overparameterized Diffusion Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Alessandro Favero, Antonio Sclocchi, Matthieu Wyart","submitted_at":"2025-05-22T17:40:08Z","abstract_excerpt":"Diffusion probabilistic models have become a cornerstone of modern generative AI, yet the mechanisms underlying their generalization remain poorly understood. In fact, if these models were perfectly minimizing their training loss, they would just generate data belonging to their training set, i.e., memorize, as empirically found in the overparameterized regime. We revisit this view by showing that, in highly overparameterized diffusion models, generalization in natural data domains is progressively achieved during training before the onset of memorization. Our results, ranging from image to la"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.16959","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/2505.16959/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":"2505.16959","created_at":"2026-07-05T12:02:46.428023+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.16959v2","created_at":"2026-07-05T12:02:46.428023+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.16959","created_at":"2026-07-05T12:02:46.428023+00:00"},{"alias_kind":"pith_short_12","alias_value":"4CFSWDKON4D6","created_at":"2026-07-05T12:02:46.428023+00:00"},{"alias_kind":"pith_short_16","alias_value":"4CFSWDKON4D6UJCE","created_at":"2026-07-05T12:02:46.428023+00:00"},{"alias_kind":"pith_short_8","alias_value":"4CFSWDKO","created_at":"2026-07-05T12:02:46.428023+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"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":23,"is_internal_anchor":true},{"citing_arxiv_id":"2606.09718","citing_title":"Evaluating the Representation Space of Diffusion Models via Self-Supervised Principles","ref_index":78,"is_internal_anchor":false},{"citing_arxiv_id":"2605.27734","citing_title":"Learn from your own latents and not from tokens: A sample-complexity theory","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21402","citing_title":"Memorisation, convergence and generalisation in generative models","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10647","citing_title":"diffGHOST: Diffusion based Generative Hedged Oblivious Synthetic Trajectories","ref_index":11,"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":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06077","citing_title":"Understanding diffusion models requires rethinking (again) generalization","ref_index":24,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4CFSWDKON4D6UJCEQ44TA567N3","json":"https://pith.science/pith/4CFSWDKON4D6UJCEQ44TA567N3.json","graph_json":"https://pith.science/api/pith-number/4CFSWDKON4D6UJCEQ44TA567N3/graph.json","events_json":"https://pith.science/api/pith-number/4CFSWDKON4D6UJCEQ44TA567N3/events.json","paper":"https://pith.science/paper/4CFSWDKO"},"agent_actions":{"view_html":"https://pith.science/pith/4CFSWDKON4D6UJCEQ44TA567N3","download_json":"https://pith.science/pith/4CFSWDKON4D6UJCEQ44TA567N3.json","view_paper":"https://pith.science/paper/4CFSWDKO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.16959&json=true","fetch_graph":"https://pith.science/api/pith-number/4CFSWDKON4D6UJCEQ44TA567N3/graph.json","fetch_events":"https://pith.science/api/pith-number/4CFSWDKON4D6UJCEQ44TA567N3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4CFSWDKON4D6UJCEQ44TA567N3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4CFSWDKON4D6UJCEQ44TA567N3/action/storage_attestation","attest_author":"https://pith.science/pith/4CFSWDKON4D6UJCEQ44TA567N3/action/author_attestation","sign_citation":"https://pith.science/pith/4CFSWDKON4D6UJCEQ44TA567N3/action/citation_signature","submit_replication":"https://pith.science/pith/4CFSWDKON4D6UJCEQ44TA567N3/action/replication_record"}},"created_at":"2026-07-05T12:02:46.428023+00:00","updated_at":"2026-07-05T12:02:46.428023+00:00"}