{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:TY6IWVMNTACLDREOKGQLN3QFXH","short_pith_number":"pith:TY6IWVMN","schema_version":"1.0","canonical_sha256":"9e3c8b558d9804b1c48e51a0b6ee05b9c29e609c7168f590ccdc559b63263044","source":{"kind":"arxiv","id":"2606.07597","version":1},"attestation_state":"computed","paper":{"title":"Repetition Mismatch: Why Data Mixture Experiments Don't Scale and How to Fix Them","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Kevin Zhou, Kris Cao, Lisa Alazraki, Marek Rei","submitted_at":"2026-05-29T06:08:57Z","abstract_excerpt":"Pre-training data mixtures are commonly tuned by running small-scale experiments and extrapolating to the target training budget. When high-quality data is scarce and must be repeated, this extrapolation frequently fails, but the source of the failure has not been isolated. We show that a primary culprit is a repetition mismatch: because high-quality datasets are small, their repetition rate changes as the training budget grows, shifting the optimal mixture in ways that small-scale proxy experiments do not anticipate. A subsampling procedure that matches the target repetition rate controls for"},"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":"2606.07597","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-29T06:08:57Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"b4127ebf0ad45b3ffb3275b3ac4da3222ee4c7dd6fd81ad4273e3108e8eaed72","abstract_canon_sha256":"f1c490f07a00338794ab20ed84009d8a613bfd5b6eac09aa78ccbda1730980ec"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-09T00:04:44.417221Z","signature_b64":"xHgFlrn4Zwk8Vxj1kPh+bGC/wrpcwn3swVb9LAOEtzVMyDMWfMDvuZo3iNKtlD6WTdTisNyvTjclwdHnGM/qAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9e3c8b558d9804b1c48e51a0b6ee05b9c29e609c7168f590ccdc559b63263044","last_reissued_at":"2026-06-09T00:04:44.416823Z","signature_status":"signed_v1","first_computed_at":"2026-06-09T00:04:44.416823Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Repetition Mismatch: Why Data Mixture Experiments Don't Scale and How to Fix Them","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Kevin Zhou, Kris Cao, Lisa Alazraki, Marek Rei","submitted_at":"2026-05-29T06:08:57Z","abstract_excerpt":"Pre-training data mixtures are commonly tuned by running small-scale experiments and extrapolating to the target training budget. When high-quality data is scarce and must be repeated, this extrapolation frequently fails, but the source of the failure has not been isolated. We show that a primary culprit is a repetition mismatch: because high-quality datasets are small, their repetition rate changes as the training budget grows, shifting the optimal mixture in ways that small-scale proxy experiments do not anticipate. A subsampling procedure that matches the target repetition rate controls for"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.07597","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/2606.07597/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":"2606.07597","created_at":"2026-06-09T00:04:44.416879+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.07597v1","created_at":"2026-06-09T00:04:44.416879+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.07597","created_at":"2026-06-09T00:04:44.416879+00:00"},{"alias_kind":"pith_short_12","alias_value":"TY6IWVMNTACL","created_at":"2026-06-09T00:04:44.416879+00:00"},{"alias_kind":"pith_short_16","alias_value":"TY6IWVMNTACLDREO","created_at":"2026-06-09T00:04:44.416879+00:00"},{"alias_kind":"pith_short_8","alias_value":"TY6IWVMN","created_at":"2026-06-09T00:04:44.416879+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/TY6IWVMNTACLDREOKGQLN3QFXH","json":"https://pith.science/pith/TY6IWVMNTACLDREOKGQLN3QFXH.json","graph_json":"https://pith.science/api/pith-number/TY6IWVMNTACLDREOKGQLN3QFXH/graph.json","events_json":"https://pith.science/api/pith-number/TY6IWVMNTACLDREOKGQLN3QFXH/events.json","paper":"https://pith.science/paper/TY6IWVMN"},"agent_actions":{"view_html":"https://pith.science/pith/TY6IWVMNTACLDREOKGQLN3QFXH","download_json":"https://pith.science/pith/TY6IWVMNTACLDREOKGQLN3QFXH.json","view_paper":"https://pith.science/paper/TY6IWVMN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.07597&json=true","fetch_graph":"https://pith.science/api/pith-number/TY6IWVMNTACLDREOKGQLN3QFXH/graph.json","fetch_events":"https://pith.science/api/pith-number/TY6IWVMNTACLDREOKGQLN3QFXH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TY6IWVMNTACLDREOKGQLN3QFXH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TY6IWVMNTACLDREOKGQLN3QFXH/action/storage_attestation","attest_author":"https://pith.science/pith/TY6IWVMNTACLDREOKGQLN3QFXH/action/author_attestation","sign_citation":"https://pith.science/pith/TY6IWVMNTACLDREOKGQLN3QFXH/action/citation_signature","submit_replication":"https://pith.science/pith/TY6IWVMNTACLDREOKGQLN3QFXH/action/replication_record"}},"created_at":"2026-06-09T00:04:44.416879+00:00","updated_at":"2026-06-09T00:04:44.416879+00:00"}