{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:JAQKJF4G5SVFVQV2LR3DZKKHDI","short_pith_number":"pith:JAQKJF4G","schema_version":"1.0","canonical_sha256":"4820a49786ecaa5ac2ba5c763ca9471a27248c600862d4b459ebf16dc8c42ed4","source":{"kind":"arxiv","id":"2506.17989","version":1},"attestation_state":"computed","paper":{"title":"Data Curation Matters: Model Collapse and Spurious Shift Performance Prediction from Training on Uncurated Text Embeddings","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Lucas Mattioli, Martin Gonzalez, Sabrina Chaouche, Youness Ait Hadichou","submitted_at":"2025-06-22T11:01:41Z","abstract_excerpt":"Training models on uncurated Text Embeddings (TEs) derived from raw tabular data can lead to a severe failure mode known as model collapse, where predictions converge to a single class regardless of input. By comparing models trained with identical hyper-parameter configurations on both raw tabular data and their TE-derived counterparts, we find that collapse is a consistent failure mode in the latter setting. We introduce a set of metrics that capture the extent of model collapse, offering a new perspective on TE quality as a proxy for data curation. Our results reveal that TE alone does not "},"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":"2506.17989","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-22T11:01:41Z","cross_cats_sorted":[],"title_canon_sha256":"7fa9134de93ca46f57cd54b2f5c7ce462a1e636dbb48e0b91f3414442a44461d","abstract_canon_sha256":"6dd08fb9bdcbd76ac5d5d4a30463c602e07b5c49da0a678a4a11a2679fe9e393"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:25:26.658976Z","signature_b64":"GykT9O0gx91bpxtnXMu2rWaAKbIIuzlZmpZ9eDGw6TMiHGrYhthLJ8dmSx3fnjBwg066fHr2IMuCXBo9728OBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4820a49786ecaa5ac2ba5c763ca9471a27248c600862d4b459ebf16dc8c42ed4","last_reissued_at":"2026-07-05T11:25:26.658494Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:25:26.658494Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Data Curation Matters: Model Collapse and Spurious Shift Performance Prediction from Training on Uncurated Text Embeddings","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Lucas Mattioli, Martin Gonzalez, Sabrina Chaouche, Youness Ait Hadichou","submitted_at":"2025-06-22T11:01:41Z","abstract_excerpt":"Training models on uncurated Text Embeddings (TEs) derived from raw tabular data can lead to a severe failure mode known as model collapse, where predictions converge to a single class regardless of input. By comparing models trained with identical hyper-parameter configurations on both raw tabular data and their TE-derived counterparts, we find that collapse is a consistent failure mode in the latter setting. We introduce a set of metrics that capture the extent of model collapse, offering a new perspective on TE quality as a proxy for data curation. Our results reveal that TE alone does not "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.17989","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/2506.17989/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":"2506.17989","created_at":"2026-07-05T11:25:26.658552+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.17989v1","created_at":"2026-07-05T11:25:26.658552+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.17989","created_at":"2026-07-05T11:25:26.658552+00:00"},{"alias_kind":"pith_short_12","alias_value":"JAQKJF4G5SVF","created_at":"2026-07-05T11:25:26.658552+00:00"},{"alias_kind":"pith_short_16","alias_value":"JAQKJF4G5SVFVQV2","created_at":"2026-07-05T11:25:26.658552+00:00"},{"alias_kind":"pith_short_8","alias_value":"JAQKJF4G","created_at":"2026-07-05T11:25:26.658552+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JAQKJF4G5SVFVQV2LR3DZKKHDI","json":"https://pith.science/pith/JAQKJF4G5SVFVQV2LR3DZKKHDI.json","graph_json":"https://pith.science/api/pith-number/JAQKJF4G5SVFVQV2LR3DZKKHDI/graph.json","events_json":"https://pith.science/api/pith-number/JAQKJF4G5SVFVQV2LR3DZKKHDI/events.json","paper":"https://pith.science/paper/JAQKJF4G"},"agent_actions":{"view_html":"https://pith.science/pith/JAQKJF4G5SVFVQV2LR3DZKKHDI","download_json":"https://pith.science/pith/JAQKJF4G5SVFVQV2LR3DZKKHDI.json","view_paper":"https://pith.science/paper/JAQKJF4G","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.17989&json=true","fetch_graph":"https://pith.science/api/pith-number/JAQKJF4G5SVFVQV2LR3DZKKHDI/graph.json","fetch_events":"https://pith.science/api/pith-number/JAQKJF4G5SVFVQV2LR3DZKKHDI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JAQKJF4G5SVFVQV2LR3DZKKHDI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JAQKJF4G5SVFVQV2LR3DZKKHDI/action/storage_attestation","attest_author":"https://pith.science/pith/JAQKJF4G5SVFVQV2LR3DZKKHDI/action/author_attestation","sign_citation":"https://pith.science/pith/JAQKJF4G5SVFVQV2LR3DZKKHDI/action/citation_signature","submit_replication":"https://pith.science/pith/JAQKJF4G5SVFVQV2LR3DZKKHDI/action/replication_record"}},"created_at":"2026-07-05T11:25:26.658552+00:00","updated_at":"2026-07-05T11:25:26.658552+00:00"}