{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:IBNWRTBGZCMTOMJ5O2U67BX2EM","short_pith_number":"pith:IBNWRTBG","schema_version":"1.0","canonical_sha256":"405b68cc26c89937313d76a9ef86fa2333d3df6febd848b00cada4dce765cd22","source":{"kind":"arxiv","id":"2410.16713","version":4},"attestation_state":"computed","paper":{"title":"Collapse or Thrive? Perils and Promises of Synthetic Data in a Self-Generating World","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Apratim Dey, David L. Donoho, Joshua Kazdan, Matthias Gerstgrasser, Rafael Rafailov, Rylan Schaeffer, Sanmi Koyejo","submitted_at":"2024-10-22T05:49:24Z","abstract_excerpt":"What happens when generative machine learning models are pretrained on web-scale datasets containing data generated by earlier models? Some prior work warns of \"model collapse\" as the web is overwhelmed by synthetic data; other work suggests the problem can be contained (i.e. collapse can be avoided) by managing how available data are used in pretraining. In this paper, we report experiments on three ways of using data (training-workflows), across three generative model task-settings (multivariate Gaussian estimation, kernel density estimation, and language-model fine-tuning) to further confir"},"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":"2410.16713","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-22T05:49:24Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"a030a9130daceadc32aca50f50afa280954ff6ac4ce12854359ef27d93310d38","abstract_canon_sha256":"8c57cfbf9431e29cb3275dd3040705f8c603657ea11d99975494fdc22b9c515c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:33:06.026956Z","signature_b64":"FPGYFRaM67yB2lhbA02lBqWHisWMYz/+jA3J0PGzxYsdrm6tj+RyyzeB1hPnktfVPw46k+LJrGv1A+m2hJP2Ag==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"405b68cc26c89937313d76a9ef86fa2333d3df6febd848b00cada4dce765cd22","last_reissued_at":"2026-07-05T10:33:06.025943Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:33:06.025943Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Collapse or Thrive? Perils and Promises of Synthetic Data in a Self-Generating World","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Apratim Dey, David L. Donoho, Joshua Kazdan, Matthias Gerstgrasser, Rafael Rafailov, Rylan Schaeffer, Sanmi Koyejo","submitted_at":"2024-10-22T05:49:24Z","abstract_excerpt":"What happens when generative machine learning models are pretrained on web-scale datasets containing data generated by earlier models? Some prior work warns of \"model collapse\" as the web is overwhelmed by synthetic data; other work suggests the problem can be contained (i.e. collapse can be avoided) by managing how available data are used in pretraining. In this paper, we report experiments on three ways of using data (training-workflows), across three generative model task-settings (multivariate Gaussian estimation, kernel density estimation, and language-model fine-tuning) to further confir"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.16713","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/2410.16713/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":"2410.16713","created_at":"2026-07-05T10:33:06.026070+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.16713v4","created_at":"2026-07-05T10:33:06.026070+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.16713","created_at":"2026-07-05T10:33:06.026070+00:00"},{"alias_kind":"pith_short_12","alias_value":"IBNWRTBGZCMT","created_at":"2026-07-05T10:33:06.026070+00:00"},{"alias_kind":"pith_short_16","alias_value":"IBNWRTBGZCMTOMJ5","created_at":"2026-07-05T10:33:06.026070+00:00"},{"alias_kind":"pith_short_8","alias_value":"IBNWRTBG","created_at":"2026-07-05T10:33:06.026070+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.08252","citing_title":"AutoPersonas: A Multi-Timescale Loop Engine for Open-Ended Persona Evolution","ref_index":6,"is_internal_anchor":true},{"citing_arxiv_id":"2606.02339","citing_title":"Entropy Minimization without Model Collapse: Mitigating Prediction Bias in Medical Imaging","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2602.17753","citing_title":"The 2025 AI Agent Index: Documenting Technical and Safety Features of Deployed Agentic AI Systems","ref_index":69,"is_internal_anchor":false},{"citing_arxiv_id":"2605.04127","citing_title":"Position: the Stochastic Parrot in the Coal Mine. Model Collapse is a Threat to Low-Resource Communities","ref_index":17,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IBNWRTBGZCMTOMJ5O2U67BX2EM","json":"https://pith.science/pith/IBNWRTBGZCMTOMJ5O2U67BX2EM.json","graph_json":"https://pith.science/api/pith-number/IBNWRTBGZCMTOMJ5O2U67BX2EM/graph.json","events_json":"https://pith.science/api/pith-number/IBNWRTBGZCMTOMJ5O2U67BX2EM/events.json","paper":"https://pith.science/paper/IBNWRTBG"},"agent_actions":{"view_html":"https://pith.science/pith/IBNWRTBGZCMTOMJ5O2U67BX2EM","download_json":"https://pith.science/pith/IBNWRTBGZCMTOMJ5O2U67BX2EM.json","view_paper":"https://pith.science/paper/IBNWRTBG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.16713&json=true","fetch_graph":"https://pith.science/api/pith-number/IBNWRTBGZCMTOMJ5O2U67BX2EM/graph.json","fetch_events":"https://pith.science/api/pith-number/IBNWRTBGZCMTOMJ5O2U67BX2EM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IBNWRTBGZCMTOMJ5O2U67BX2EM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IBNWRTBGZCMTOMJ5O2U67BX2EM/action/storage_attestation","attest_author":"https://pith.science/pith/IBNWRTBGZCMTOMJ5O2U67BX2EM/action/author_attestation","sign_citation":"https://pith.science/pith/IBNWRTBGZCMTOMJ5O2U67BX2EM/action/citation_signature","submit_replication":"https://pith.science/pith/IBNWRTBGZCMTOMJ5O2U67BX2EM/action/replication_record"}},"created_at":"2026-07-05T10:33:06.026070+00:00","updated_at":"2026-07-05T10:33:06.026070+00:00"}