{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:RGZXJO2T3HKX2JQVP4ZOXQJJR5","short_pith_number":"pith:RGZXJO2T","schema_version":"1.0","canonical_sha256":"89b374bb53d9d57d26157f32ebc1298f7b9de6def47520f610193141c8f7aea0","source":{"kind":"arxiv","id":"2404.01413","version":2},"attestation_state":"computed","paper":{"title":"Is Model Collapse Inevitable? Breaking the Curse of Recursion by Accumulating Real and Synthetic Data","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.ET","stat.ML"],"primary_cat":"cs.LG","authors_text":"Andrey Gromov, Apratim Dey, Daniel A. Roberts, David L. Donoho, Dhruv Pai, Diyi Yang, Henry Sleight, John Hughes, Matthias Gerstgrasser, Rafael Rafailov, Rajashree Agrawal, Rylan Schaeffer, Sanmi Koyejo, Tomasz Korbak","submitted_at":"2024-04-01T18:31:24Z","abstract_excerpt":"The proliferation of generative models, combined with pretraining on web-scale data, raises a timely question: what happens when these models are trained on their own generated outputs? Recent investigations into model-data feedback loops proposed that such loops would lead to a phenomenon termed model collapse, under which performance progressively degrades with each model-data feedback iteration until fitted models become useless. However, those studies largely assumed that new data replace old data over time, where an arguably more realistic assumption is that data accumulate over time. In "},"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":"2404.01413","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-04-01T18:31:24Z","cross_cats_sorted":["cs.AI","cs.CL","cs.ET","stat.ML"],"title_canon_sha256":"6294c8fe5521f37310c73f4c9184935437a58a42660f022557eec9cb32d44001","abstract_canon_sha256":"eec5a946adfcfe1d88cd8071e870ba1df5a88c8e28e5f573fc87be87ae846cee"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:13:31.450481Z","signature_b64":"znGcbLAnwQ0xyWiJuqWlCXL8d4z7O9CHjcwca06D7ULNvAKIynpodcexXmiZX1hho9QmPogsTQc7FiUHpufxBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"89b374bb53d9d57d26157f32ebc1298f7b9de6def47520f610193141c8f7aea0","last_reissued_at":"2026-07-05T08:13:31.449990Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:13:31.449990Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Is Model Collapse Inevitable? Breaking the Curse of Recursion by Accumulating Real and Synthetic Data","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.ET","stat.ML"],"primary_cat":"cs.LG","authors_text":"Andrey Gromov, Apratim Dey, Daniel A. Roberts, David L. Donoho, Dhruv Pai, Diyi Yang, Henry Sleight, John Hughes, Matthias Gerstgrasser, Rafael Rafailov, Rajashree Agrawal, Rylan Schaeffer, Sanmi Koyejo, Tomasz Korbak","submitted_at":"2024-04-01T18:31:24Z","abstract_excerpt":"The proliferation of generative models, combined with pretraining on web-scale data, raises a timely question: what happens when these models are trained on their own generated outputs? Recent investigations into model-data feedback loops proposed that such loops would lead to a phenomenon termed model collapse, under which performance progressively degrades with each model-data feedback iteration until fitted models become useless. However, those studies largely assumed that new data replace old data over time, where an arguably more realistic assumption is that data accumulate over time. In "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.01413","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/2404.01413/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":"2404.01413","created_at":"2026-07-05T08:13:31.450048+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.01413v2","created_at":"2026-07-05T08:13:31.450048+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.01413","created_at":"2026-07-05T08:13:31.450048+00:00"},{"alias_kind":"pith_short_12","alias_value":"RGZXJO2T3HKX","created_at":"2026-07-05T08:13:31.450048+00:00"},{"alias_kind":"pith_short_16","alias_value":"RGZXJO2T3HKX2JQV","created_at":"2026-07-05T08:13:31.450048+00:00"},{"alias_kind":"pith_short_8","alias_value":"RGZXJO2T","created_at":"2026-07-05T08:13:31.450048+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":12,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.26492","citing_title":"Elias in the Lighthouse, Again? Diagnosing Low Diversity in LLM Stories","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2504.12501","citing_title":"Reinforcement Learning from Human Feedback","ref_index":274,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20279","citing_title":"The Economics of Model Collapse: Equilibrium, Welfare, and Optimal Provenance Subsidies in Synthetic Data Markets","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2508.16168","citing_title":"FuXi-TC: A generative framework integrating deep learning and physics-based models for improved tropical cyclone forecasts","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2604.08554","citing_title":"Drift and selection in LLM text ecosystems","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2604.26855","citing_title":"Cognitive Atrophy and Systemic Collapse in AI-Dependent Software Engineering","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25110","citing_title":"Knowledge Distillation Must Account for What It Loses","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25110","citing_title":"Knowledge Distillation Must Account for What It Loses","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07724","citing_title":"Curated Synthetic Data Doesn't Have to Collapse: A Theoretical Study of Generative Retraining with Pluralistic Preferences","ref_index":78,"is_internal_anchor":false},{"citing_arxiv_id":"2604.26965","citing_title":"The Impact of AI-Generated Text on the Internet","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2604.15786","citing_title":"Filter Babel: The Challenge of Synthetic Media to Authenticity and Common Ground in AI-Mediated Communication","ref_index":19,"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":11,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RGZXJO2T3HKX2JQVP4ZOXQJJR5","json":"https://pith.science/pith/RGZXJO2T3HKX2JQVP4ZOXQJJR5.json","graph_json":"https://pith.science/api/pith-number/RGZXJO2T3HKX2JQVP4ZOXQJJR5/graph.json","events_json":"https://pith.science/api/pith-number/RGZXJO2T3HKX2JQVP4ZOXQJJR5/events.json","paper":"https://pith.science/paper/RGZXJO2T"},"agent_actions":{"view_html":"https://pith.science/pith/RGZXJO2T3HKX2JQVP4ZOXQJJR5","download_json":"https://pith.science/pith/RGZXJO2T3HKX2JQVP4ZOXQJJR5.json","view_paper":"https://pith.science/paper/RGZXJO2T","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.01413&json=true","fetch_graph":"https://pith.science/api/pith-number/RGZXJO2T3HKX2JQVP4ZOXQJJR5/graph.json","fetch_events":"https://pith.science/api/pith-number/RGZXJO2T3HKX2JQVP4ZOXQJJR5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RGZXJO2T3HKX2JQVP4ZOXQJJR5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RGZXJO2T3HKX2JQVP4ZOXQJJR5/action/storage_attestation","attest_author":"https://pith.science/pith/RGZXJO2T3HKX2JQVP4ZOXQJJR5/action/author_attestation","sign_citation":"https://pith.science/pith/RGZXJO2T3HKX2JQVP4ZOXQJJR5/action/citation_signature","submit_replication":"https://pith.science/pith/RGZXJO2T3HKX2JQVP4ZOXQJJR5/action/replication_record"}},"created_at":"2026-07-05T08:13:31.450048+00:00","updated_at":"2026-07-05T08:13:31.450048+00:00"}