{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:3WGJGZ47ZPBLAZ2IICLMN67IEE","short_pith_number":"pith:3WGJGZ47","schema_version":"1.0","canonical_sha256":"dd8c93679fcbc2b067484096c6fbe82114e8acc561cbcd1c39d33c514b9da173","source":{"kind":"arxiv","id":"2410.07064","version":2},"attestation_state":"computed","paper":{"title":"Data Selection via Optimal Control for Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Furu Wei, Hongning Wang, Li Dong, Minlie Huang, Qingxiu Dong, Yaru Hao, Yuxian Gu","submitted_at":"2024-10-09T17:06:57Z","abstract_excerpt":"This work investigates the selection of high-quality pre-training data from massive corpora to enhance LMs' capabilities for downstream usage. We formulate data selection as a generalized Optimal Control problem, which can be solved theoretically by Pontryagin's Maximum Principle (PMP), yielding a set of necessary conditions that characterize the relationship between optimal data selection and LM training dynamics. Based on these theoretical results, we introduce PMP-based Data Selection (PDS), a framework that approximates optimal data selection by solving the PMP conditions. In our experimen"},"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":"2410.07064","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-10-09T17:06:57Z","cross_cats_sorted":[],"title_canon_sha256":"4b120f2d81b7f1a4250cd9e2096cfa2ccc1cc0e44f8cc0efbe81b267d891361e","abstract_canon_sha256":"21cda24e38800e411b4092eb5b9c164411473c0be86dc2aa03db0a30abde05b8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:34:11.388562Z","signature_b64":"xd0lG2cxiIx2oGE7IYusyY6E7YjMwdnVSZvx0rtbarOY7YB7Clkd9keXqLyWw79ZW0in+WzCOkUwSIeZu+DBDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"dd8c93679fcbc2b067484096c6fbe82114e8acc561cbcd1c39d33c514b9da173","last_reissued_at":"2026-07-05T10:34:11.387516Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:34:11.387516Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Data Selection via Optimal Control for Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Furu Wei, Hongning Wang, Li Dong, Minlie Huang, Qingxiu Dong, Yaru Hao, Yuxian Gu","submitted_at":"2024-10-09T17:06:57Z","abstract_excerpt":"This work investigates the selection of high-quality pre-training data from massive corpora to enhance LMs' capabilities for downstream usage. We formulate data selection as a generalized Optimal Control problem, which can be solved theoretically by Pontryagin's Maximum Principle (PMP), yielding a set of necessary conditions that characterize the relationship between optimal data selection and LM training dynamics. Based on these theoretical results, we introduce PMP-based Data Selection (PDS), a framework that approximates optimal data selection by solving the PMP conditions. In our experimen"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.07064","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/2410.07064/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":"2410.07064","created_at":"2026-07-05T10:34:11.387652+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.07064v2","created_at":"2026-07-05T10:34:11.387652+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.07064","created_at":"2026-07-05T10:34:11.387652+00:00"},{"alias_kind":"pith_short_12","alias_value":"3WGJGZ47ZPBL","created_at":"2026-07-05T10:34:11.387652+00:00"},{"alias_kind":"pith_short_16","alias_value":"3WGJGZ47ZPBLAZ2I","created_at":"2026-07-05T10:34:11.387652+00:00"},{"alias_kind":"pith_short_8","alias_value":"3WGJGZ47","created_at":"2026-07-05T10:34:11.387652+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.00251","citing_title":"Capability Self-Assessment: Teaching LLMs to Know Their Limits","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2412.04468","citing_title":"NVILA: Efficient Frontier Visual Language Models","ref_index":28,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3WGJGZ47ZPBLAZ2IICLMN67IEE","json":"https://pith.science/pith/3WGJGZ47ZPBLAZ2IICLMN67IEE.json","graph_json":"https://pith.science/api/pith-number/3WGJGZ47ZPBLAZ2IICLMN67IEE/graph.json","events_json":"https://pith.science/api/pith-number/3WGJGZ47ZPBLAZ2IICLMN67IEE/events.json","paper":"https://pith.science/paper/3WGJGZ47"},"agent_actions":{"view_html":"https://pith.science/pith/3WGJGZ47ZPBLAZ2IICLMN67IEE","download_json":"https://pith.science/pith/3WGJGZ47ZPBLAZ2IICLMN67IEE.json","view_paper":"https://pith.science/paper/3WGJGZ47","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.07064&json=true","fetch_graph":"https://pith.science/api/pith-number/3WGJGZ47ZPBLAZ2IICLMN67IEE/graph.json","fetch_events":"https://pith.science/api/pith-number/3WGJGZ47ZPBLAZ2IICLMN67IEE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3WGJGZ47ZPBLAZ2IICLMN67IEE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3WGJGZ47ZPBLAZ2IICLMN67IEE/action/storage_attestation","attest_author":"https://pith.science/pith/3WGJGZ47ZPBLAZ2IICLMN67IEE/action/author_attestation","sign_citation":"https://pith.science/pith/3WGJGZ47ZPBLAZ2IICLMN67IEE/action/citation_signature","submit_replication":"https://pith.science/pith/3WGJGZ47ZPBLAZ2IICLMN67IEE/action/replication_record"}},"created_at":"2026-07-05T10:34:11.387652+00:00","updated_at":"2026-07-05T10:34:11.387652+00:00"}