{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:ANQRXWMAD4E7D3RVF4TG752ACE","short_pith_number":"pith:ANQRXWMA","schema_version":"1.0","canonical_sha256":"03611bd9801f09f1ee352f266ff7401120bdb61fcd65f62c14600ce2b679339b","source":{"kind":"arxiv","id":"2505.08803","version":1},"attestation_state":"computed","paper":{"title":"Multi-modal Synthetic Data Training and Model Collapse: Insights from VLMs and Diffusion Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Jesse Thomason, Mohammad Rostami, Zizhao Hu","submitted_at":"2025-05-10T22:42:29Z","abstract_excerpt":"Recent research has highlighted the risk of generative model collapse, where performance progressively degrades when continually trained on self-generated data. However, existing exploration on model collapse is limited to single, unimodal models, limiting our understanding in more realistic scenarios, such as diverse multi-modal AI agents interacting autonomously through synthetic data and continually evolving. We expand the synthetic data training and model collapse study to multi-modal vision-language generative systems, such as vision-language models (VLMs) and text-to-image diffusion mode"},"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":"2505.08803","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-10T22:42:29Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"94797d8cb20c9b33468cc9678901d111573e8bd9e7dbbf630bb64cfcd2ef5b7a","abstract_canon_sha256":"82a9ad5a31ca6ee3daefe4d20568f84fb1cac70f71458c04acb869244ff75850"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:02:44.597954Z","signature_b64":"fPUrLDti8vwG2wTazZSlkLZJP+Oohru7CHoBEM6ajL40o+Ea8uE3Aa5eOatv/AxtDw0aWDwqF5fTVxH+RbBUCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"03611bd9801f09f1ee352f266ff7401120bdb61fcd65f62c14600ce2b679339b","last_reissued_at":"2026-07-05T11:02:44.597391Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:02:44.597391Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Multi-modal Synthetic Data Training and Model Collapse: Insights from VLMs and Diffusion Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Jesse Thomason, Mohammad Rostami, Zizhao Hu","submitted_at":"2025-05-10T22:42:29Z","abstract_excerpt":"Recent research has highlighted the risk of generative model collapse, where performance progressively degrades when continually trained on self-generated data. However, existing exploration on model collapse is limited to single, unimodal models, limiting our understanding in more realistic scenarios, such as diverse multi-modal AI agents interacting autonomously through synthetic data and continually evolving. We expand the synthetic data training and model collapse study to multi-modal vision-language generative systems, such as vision-language models (VLMs) and text-to-image diffusion mode"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.08803","kind":"arxiv","version":1},"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/2505.08803/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":"2505.08803","created_at":"2026-07-05T11:02:44.597449+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.08803v1","created_at":"2026-07-05T11:02:44.597449+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.08803","created_at":"2026-07-05T11:02:44.597449+00:00"},{"alias_kind":"pith_short_12","alias_value":"ANQRXWMAD4E7","created_at":"2026-07-05T11:02:44.597449+00:00"},{"alias_kind":"pith_short_16","alias_value":"ANQRXWMAD4E7D3RV","created_at":"2026-07-05T11:02:44.597449+00:00"},{"alias_kind":"pith_short_8","alias_value":"ANQRXWMA","created_at":"2026-07-05T11:02:44.597449+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.09022","citing_title":"BlendFusion -- Scalable Synthetic Data Generation for Diffusion Model Training","ref_index":9,"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":96,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ANQRXWMAD4E7D3RVF4TG752ACE","json":"https://pith.science/pith/ANQRXWMAD4E7D3RVF4TG752ACE.json","graph_json":"https://pith.science/api/pith-number/ANQRXWMAD4E7D3RVF4TG752ACE/graph.json","events_json":"https://pith.science/api/pith-number/ANQRXWMAD4E7D3RVF4TG752ACE/events.json","paper":"https://pith.science/paper/ANQRXWMA"},"agent_actions":{"view_html":"https://pith.science/pith/ANQRXWMAD4E7D3RVF4TG752ACE","download_json":"https://pith.science/pith/ANQRXWMAD4E7D3RVF4TG752ACE.json","view_paper":"https://pith.science/paper/ANQRXWMA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.08803&json=true","fetch_graph":"https://pith.science/api/pith-number/ANQRXWMAD4E7D3RVF4TG752ACE/graph.json","fetch_events":"https://pith.science/api/pith-number/ANQRXWMAD4E7D3RVF4TG752ACE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ANQRXWMAD4E7D3RVF4TG752ACE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ANQRXWMAD4E7D3RVF4TG752ACE/action/storage_attestation","attest_author":"https://pith.science/pith/ANQRXWMAD4E7D3RVF4TG752ACE/action/author_attestation","sign_citation":"https://pith.science/pith/ANQRXWMAD4E7D3RVF4TG752ACE/action/citation_signature","submit_replication":"https://pith.science/pith/ANQRXWMAD4E7D3RVF4TG752ACE/action/replication_record"}},"created_at":"2026-07-05T11:02:44.597449+00:00","updated_at":"2026-07-05T11:02:44.597449+00:00"}