{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:ODR4PEDTAJAU2RASHTXXBLXTVO","short_pith_number":"pith:ODR4PEDT","schema_version":"1.0","canonical_sha256":"70e3c7907302414d44123cef70aef3ab96ab6b539bbaf1a4b268eaf37be93e75","source":{"kind":"arxiv","id":"2512.15011","version":3},"attestation_state":"computed","paper":{"title":"Epistemic diversity across language models mitigates knowledge collapse","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CY","cs.MA"],"primary_cat":"cs.LG","authors_text":"Damian Hodel, Jevin D. West","submitted_at":"2025-12-17T02:03:28Z","abstract_excerpt":"Artificial intelligence (AI) increasingly generates the very content used to train future AI systems. This feedback loop can degrade model quality, reduce informational diversity, and ultimately drive knowledge collapse, i.e. a degradation to a narrow and inaccurate set of ideas. We ask: to mitigate collapse, is it better to concentrate the internet's knowledge into a handful of dominant models (referred to as an AI monoculture), or to distribute it across a diverse ecosystem of models? To study the effect of diversity on model performance, we randomly segment the fixed training data across an"},"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":"2512.15011","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-12-17T02:03:28Z","cross_cats_sorted":["cs.AI","cs.CY","cs.MA"],"title_canon_sha256":"35c78c1e822ddca6add44b512f126ffe1fa5682c711b8f5bac7b5406174e41e9","abstract_canon_sha256":"20110c226ae5cac6accd14d779a177155d1471f00a2831bd515c662e34ad2a36"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"70e3c7907302414d44123cef70aef3ab96ab6b539bbaf1a4b268eaf37be93e75","last_reissued_at":"2026-07-31T01:08:56.688143Z","signature_status":"unsigned_v0","first_computed_at":"2026-07-31T01:08:56.688143Z"},"graph_snapshot":{"paper":{"title":"Epistemic diversity across language models mitigates knowledge collapse","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CY","cs.MA"],"primary_cat":"cs.LG","authors_text":"Damian Hodel, Jevin D. West","submitted_at":"2025-12-17T02:03:28Z","abstract_excerpt":"Artificial intelligence (AI) increasingly generates the very content used to train future AI systems. This feedback loop can degrade model quality, reduce informational diversity, and ultimately drive knowledge collapse, i.e. a degradation to a narrow and inaccurate set of ideas. We ask: to mitigate collapse, is it better to concentrate the internet's knowledge into a handful of dominant models (referred to as an AI monoculture), or to distribute it across a diverse ecosystem of models? To study the effect of diversity on model performance, we randomly segment the fixed training data across an"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2512.15011","kind":"arxiv","version":3},"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/2512.15011/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":"2512.15011","created_at":"2026-07-31T01:08:56.691201+00:00"},{"alias_kind":"arxiv_version","alias_value":"2512.15011v3","created_at":"2026-07-31T01:08:56.691201+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2512.15011","created_at":"2026-07-31T01:08:56.691201+00:00"},{"alias_kind":"pith_short_12","alias_value":"ODR4PEDTAJAU","created_at":"2026-07-31T01:08:56.691201+00:00"},{"alias_kind":"pith_short_16","alias_value":"ODR4PEDTAJAU2RAS","created_at":"2026-07-31T01:08:56.691201+00:00"},{"alias_kind":"pith_short_8","alias_value":"ODR4PEDT","created_at":"2026-07-31T01:08:56.691201+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2606.02632","citing_title":"Position: Prioritize Identifying Structure, Not Complex Models, for Scientific Discovery","ref_index":50,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ODR4PEDTAJAU2RASHTXXBLXTVO","json":"https://pith.science/pith/ODR4PEDTAJAU2RASHTXXBLXTVO.json","graph_json":"https://pith.science/api/pith-number/ODR4PEDTAJAU2RASHTXXBLXTVO/graph.json","events_json":"https://pith.science/api/pith-number/ODR4PEDTAJAU2RASHTXXBLXTVO/events.json","paper":"https://pith.science/paper/ODR4PEDT"},"agent_actions":{"view_html":"https://pith.science/pith/ODR4PEDTAJAU2RASHTXXBLXTVO","download_json":"https://pith.science/pith/ODR4PEDTAJAU2RASHTXXBLXTVO.json","view_paper":"https://pith.science/paper/ODR4PEDT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2512.15011&json=true","fetch_graph":"https://pith.science/api/pith-number/ODR4PEDTAJAU2RASHTXXBLXTVO/graph.json","fetch_events":"https://pith.science/api/pith-number/ODR4PEDTAJAU2RASHTXXBLXTVO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ODR4PEDTAJAU2RASHTXXBLXTVO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ODR4PEDTAJAU2RASHTXXBLXTVO/action/storage_attestation","attest_author":"https://pith.science/pith/ODR4PEDTAJAU2RASHTXXBLXTVO/action/author_attestation","sign_citation":"https://pith.science/pith/ODR4PEDTAJAU2RASHTXXBLXTVO/action/citation_signature","submit_replication":"https://pith.science/pith/ODR4PEDTAJAU2RASHTXXBLXTVO/action/replication_record"}},"created_at":"2026-07-31T01:08:56.691201+00:00","updated_at":"2026-07-31T01:08:56.691201+00:00"}