{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:BL6LQHCWF6WXXQFAQ5WRQBD4GR","short_pith_number":"pith:BL6LQHCW","schema_version":"1.0","canonical_sha256":"0afcb81c562fad7bc0a0876d18047c344cecb9b178be6f497a24405e72dc2aa9","source":{"kind":"arxiv","id":"2606.25086","version":1},"attestation_state":"computed","paper":{"title":"Training for the Model You Return: Improving Optimization for Iterate-Averaged Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Adam Block, Kwok Chun Au","submitted_at":"2026-06-23T18:47:40Z","abstract_excerpt":"Many modern Language Model (LM) pipelines return an averaged model, such as an exponential moving average of the training iterates, rather than the final iterate itself. This raises a fundamental question: given that we will return an iterate average, how should we change training to improve the performance of this average? We study this question by formulating optimizer design for the iterate-average estimator as an optimal-control problem. In a continuous-time stochastic quadratic model, we solve for the control strategy that minimizes the error of the returned average subject to a penalty o"},"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":"2606.25086","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-23T18:47:40Z","cross_cats_sorted":["cs.AI","stat.ML"],"title_canon_sha256":"e4788679d8ecefb2304c280b849f8b467088730106edb40f2bf4789fee62bab5","abstract_canon_sha256":"37c16f984819feec3e35f64dd4ca70ee22eb80a29a9455dea4484e26faa97636"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-25T00:18:17.356127Z","signature_b64":"+HOpg7wwmBgGSnvkJLwX2QLsvw1iJzoeT0NHAA9byeDP2wYeUAnqgjuV793N+5ja6JGh0+8SMlzlvFW6tb7JBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0afcb81c562fad7bc0a0876d18047c344cecb9b178be6f497a24405e72dc2aa9","last_reissued_at":"2026-06-25T00:18:17.355775Z","signature_status":"signed_v1","first_computed_at":"2026-06-25T00:18:17.355775Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Training for the Model You Return: Improving Optimization for Iterate-Averaged Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Adam Block, Kwok Chun Au","submitted_at":"2026-06-23T18:47:40Z","abstract_excerpt":"Many modern Language Model (LM) pipelines return an averaged model, such as an exponential moving average of the training iterates, rather than the final iterate itself. This raises a fundamental question: given that we will return an iterate average, how should we change training to improve the performance of this average? We study this question by formulating optimizer design for the iterate-average estimator as an optimal-control problem. In a continuous-time stochastic quadratic model, we solve for the control strategy that minimizes the error of the returned average subject to a penalty o"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.25086","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/2606.25086/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":"2606.25086","created_at":"2026-06-25T00:18:17.355835+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.25086v1","created_at":"2026-06-25T00:18:17.355835+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.25086","created_at":"2026-06-25T00:18:17.355835+00:00"},{"alias_kind":"pith_short_12","alias_value":"BL6LQHCWF6WX","created_at":"2026-06-25T00:18:17.355835+00:00"},{"alias_kind":"pith_short_16","alias_value":"BL6LQHCWF6WXXQFA","created_at":"2026-06-25T00:18:17.355835+00:00"},{"alias_kind":"pith_short_8","alias_value":"BL6LQHCW","created_at":"2026-06-25T00:18:17.355835+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/BL6LQHCWF6WXXQFAQ5WRQBD4GR","json":"https://pith.science/pith/BL6LQHCWF6WXXQFAQ5WRQBD4GR.json","graph_json":"https://pith.science/api/pith-number/BL6LQHCWF6WXXQFAQ5WRQBD4GR/graph.json","events_json":"https://pith.science/api/pith-number/BL6LQHCWF6WXXQFAQ5WRQBD4GR/events.json","paper":"https://pith.science/paper/BL6LQHCW"},"agent_actions":{"view_html":"https://pith.science/pith/BL6LQHCWF6WXXQFAQ5WRQBD4GR","download_json":"https://pith.science/pith/BL6LQHCWF6WXXQFAQ5WRQBD4GR.json","view_paper":"https://pith.science/paper/BL6LQHCW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.25086&json=true","fetch_graph":"https://pith.science/api/pith-number/BL6LQHCWF6WXXQFAQ5WRQBD4GR/graph.json","fetch_events":"https://pith.science/api/pith-number/BL6LQHCWF6WXXQFAQ5WRQBD4GR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BL6LQHCWF6WXXQFAQ5WRQBD4GR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BL6LQHCWF6WXXQFAQ5WRQBD4GR/action/storage_attestation","attest_author":"https://pith.science/pith/BL6LQHCWF6WXXQFAQ5WRQBD4GR/action/author_attestation","sign_citation":"https://pith.science/pith/BL6LQHCWF6WXXQFAQ5WRQBD4GR/action/citation_signature","submit_replication":"https://pith.science/pith/BL6LQHCWF6WXXQFAQ5WRQBD4GR/action/replication_record"}},"created_at":"2026-06-25T00:18:17.355835+00:00","updated_at":"2026-06-25T00:18:17.355835+00:00"}