{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:GMKFNNXDP5RWKFLE6GFOGVXQBC","short_pith_number":"pith:GMKFNNXD","schema_version":"1.0","canonical_sha256":"331456b6e37f63651564f18ae356f0088781b6f0517184238f6b841166f3054d","source":{"kind":"arxiv","id":"2502.15950","version":1},"attestation_state":"computed","paper":{"title":"Optimizing Pre-Training Data Mixtures with Mixtures of Data Expert Models","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Alekh Agarwal, Kristina Toutanova, Lior Belenki, Tianze Shi","submitted_at":"2025-02-21T21:27:48Z","abstract_excerpt":"We propose a method to optimize language model pre-training data mixtures through efficient approximation of the cross-entropy loss corresponding to each candidate mixture via a Mixture of Data Experts (MDE). We use this approximation as a source of additional features in a regression model, trained from observations of model loss for a small number of mixtures.\n  Experiments with Transformer decoder-only language models in the range of 70M to 1B parameters on the SlimPajama dataset show that our method achieves significantly better performance than approaches that train regression models usin"},"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":"2502.15950","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-21T21:27:48Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"81d174b60ba1bed421b79ed69dab96b2325952fb842b6e881a4991ed5157b435","abstract_canon_sha256":"f0722d6de25663e7d09d87db38c7b273d1dfd257c9867c301967e7f4eb221f2e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:18:18.907665Z","signature_b64":"r/pCzeZD/u/kAm8d7Vytatl/MgsDq/XUd9Wvpfc7Fyn0xI1uAweJxAkuhQeVPSDB47L44S8zSg91GQE2WyIzAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"331456b6e37f63651564f18ae356f0088781b6f0517184238f6b841166f3054d","last_reissued_at":"2026-07-05T10:18:18.907222Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:18:18.907222Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Optimizing Pre-Training Data Mixtures with Mixtures of Data Expert Models","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Alekh Agarwal, Kristina Toutanova, Lior Belenki, Tianze Shi","submitted_at":"2025-02-21T21:27:48Z","abstract_excerpt":"We propose a method to optimize language model pre-training data mixtures through efficient approximation of the cross-entropy loss corresponding to each candidate mixture via a Mixture of Data Experts (MDE). We use this approximation as a source of additional features in a regression model, trained from observations of model loss for a small number of mixtures.\n  Experiments with Transformer decoder-only language models in the range of 70M to 1B parameters on the SlimPajama dataset show that our method achieves significantly better performance than approaches that train regression models usin"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.15950","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/2502.15950/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":"2502.15950","created_at":"2026-07-05T10:18:18.907275+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.15950v1","created_at":"2026-07-05T10:18:18.907275+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.15950","created_at":"2026-07-05T10:18:18.907275+00:00"},{"alias_kind":"pith_short_12","alias_value":"GMKFNNXDP5RW","created_at":"2026-07-05T10:18:18.907275+00:00"},{"alias_kind":"pith_short_16","alias_value":"GMKFNNXDP5RWKFLE","created_at":"2026-07-05T10:18:18.907275+00:00"},{"alias_kind":"pith_short_8","alias_value":"GMKFNNXD","created_at":"2026-07-05T10:18:18.907275+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/GMKFNNXDP5RWKFLE6GFOGVXQBC","json":"https://pith.science/pith/GMKFNNXDP5RWKFLE6GFOGVXQBC.json","graph_json":"https://pith.science/api/pith-number/GMKFNNXDP5RWKFLE6GFOGVXQBC/graph.json","events_json":"https://pith.science/api/pith-number/GMKFNNXDP5RWKFLE6GFOGVXQBC/events.json","paper":"https://pith.science/paper/GMKFNNXD"},"agent_actions":{"view_html":"https://pith.science/pith/GMKFNNXDP5RWKFLE6GFOGVXQBC","download_json":"https://pith.science/pith/GMKFNNXDP5RWKFLE6GFOGVXQBC.json","view_paper":"https://pith.science/paper/GMKFNNXD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.15950&json=true","fetch_graph":"https://pith.science/api/pith-number/GMKFNNXDP5RWKFLE6GFOGVXQBC/graph.json","fetch_events":"https://pith.science/api/pith-number/GMKFNNXDP5RWKFLE6GFOGVXQBC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GMKFNNXDP5RWKFLE6GFOGVXQBC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GMKFNNXDP5RWKFLE6GFOGVXQBC/action/storage_attestation","attest_author":"https://pith.science/pith/GMKFNNXDP5RWKFLE6GFOGVXQBC/action/author_attestation","sign_citation":"https://pith.science/pith/GMKFNNXDP5RWKFLE6GFOGVXQBC/action/citation_signature","submit_replication":"https://pith.science/pith/GMKFNNXDP5RWKFLE6GFOGVXQBC/action/replication_record"}},"created_at":"2026-07-05T10:18:18.907275+00:00","updated_at":"2026-07-05T10:18:18.907275+00:00"}