{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:QVADTLOFOY5MSCXVR3ONMQ7T3L","short_pith_number":"pith:QVADTLOF","schema_version":"1.0","canonical_sha256":"854039adc5763ac90af58edcd643f3dac5128ddd421b179183f665befa0c62fe","source":{"kind":"arxiv","id":"2505.13947","version":2},"attestation_state":"computed","paper":{"title":"A Probabilistic Perspective on Model Collapse","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Guang Cheng, Hengzhi He, Shirong Xu","submitted_at":"2025-05-20T05:25:29Z","abstract_excerpt":"In recent years, model collapse has become a critical issue in language model training, making it essential to understand the underlying mechanisms driving this phenomenon. In this paper, we investigate recursive parametric model training from a probabilistic perspective, aiming to characterize the conditions under which model collapse occurs and, crucially, how it can be mitigated. We conceptualize the recursive training process as a random walk of the model estimate, highlighting how the sample size influences the step size and how the estimation procedure determines the direction and potent"},"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":"2505.13947","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2025-05-20T05:25:29Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"b32ffa46e3e41f24d6e4233e774f63ed4a9e842032fdeb701819502048fd1296","abstract_canon_sha256":"ae13dcc5b6ddacd9b63a87889f615fec78742846945b10c8e13c151c122848b9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:07:25.124775Z","signature_b64":"sIWYbGFQ5qIMaDzkc0em+eYg/b3utNoox/JQtgt1f6sxMH9p+/9HRs49bpPaPL/rUNLVl6UsY4BDN371EwkIBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"854039adc5763ac90af58edcd643f3dac5128ddd421b179183f665befa0c62fe","last_reissued_at":"2026-07-05T11:07:25.124284Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:07:25.124284Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Probabilistic Perspective on Model Collapse","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Guang Cheng, Hengzhi He, Shirong Xu","submitted_at":"2025-05-20T05:25:29Z","abstract_excerpt":"In recent years, model collapse has become a critical issue in language model training, making it essential to understand the underlying mechanisms driving this phenomenon. In this paper, we investigate recursive parametric model training from a probabilistic perspective, aiming to characterize the conditions under which model collapse occurs and, crucially, how it can be mitigated. We conceptualize the recursive training process as a random walk of the model estimate, highlighting how the sample size influences the step size and how the estimation procedure determines the direction and potent"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.13947","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/2505.13947/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":"2505.13947","created_at":"2026-07-05T11:07:25.124355+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.13947v2","created_at":"2026-07-05T11:07:25.124355+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.13947","created_at":"2026-07-05T11:07:25.124355+00:00"},{"alias_kind":"pith_short_12","alias_value":"QVADTLOFOY5M","created_at":"2026-07-05T11:07:25.124355+00:00"},{"alias_kind":"pith_short_16","alias_value":"QVADTLOFOY5MSCXV","created_at":"2026-07-05T11:07:25.124355+00:00"},{"alias_kind":"pith_short_8","alias_value":"QVADTLOF","created_at":"2026-07-05T11:07:25.124355+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"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":103,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QVADTLOFOY5MSCXVR3ONMQ7T3L","json":"https://pith.science/pith/QVADTLOFOY5MSCXVR3ONMQ7T3L.json","graph_json":"https://pith.science/api/pith-number/QVADTLOFOY5MSCXVR3ONMQ7T3L/graph.json","events_json":"https://pith.science/api/pith-number/QVADTLOFOY5MSCXVR3ONMQ7T3L/events.json","paper":"https://pith.science/paper/QVADTLOF"},"agent_actions":{"view_html":"https://pith.science/pith/QVADTLOFOY5MSCXVR3ONMQ7T3L","download_json":"https://pith.science/pith/QVADTLOFOY5MSCXVR3ONMQ7T3L.json","view_paper":"https://pith.science/paper/QVADTLOF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.13947&json=true","fetch_graph":"https://pith.science/api/pith-number/QVADTLOFOY5MSCXVR3ONMQ7T3L/graph.json","fetch_events":"https://pith.science/api/pith-number/QVADTLOFOY5MSCXVR3ONMQ7T3L/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QVADTLOFOY5MSCXVR3ONMQ7T3L/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QVADTLOFOY5MSCXVR3ONMQ7T3L/action/storage_attestation","attest_author":"https://pith.science/pith/QVADTLOFOY5MSCXVR3ONMQ7T3L/action/author_attestation","sign_citation":"https://pith.science/pith/QVADTLOFOY5MSCXVR3ONMQ7T3L/action/citation_signature","submit_replication":"https://pith.science/pith/QVADTLOFOY5MSCXVR3ONMQ7T3L/action/replication_record"}},"created_at":"2026-07-05T11:07:25.124355+00:00","updated_at":"2026-07-05T11:07:25.124355+00:00"}