{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:EWX7F7ZXY7P26FXDPZQN3RICMB","short_pith_number":"pith:EWX7F7ZX","schema_version":"1.0","canonical_sha256":"25aff2ff37c7dfaf16e37e60ddc50260782d5b1b92dc156bdabbb4c8ea05de55","source":{"kind":"arxiv","id":"2505.06699","version":3},"attestation_state":"computed","paper":{"title":"Model Steering: Learning with a Reference Model Improves Generalization Bounds and Scaling Laws","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CV","stat.ML"],"primary_cat":"cs.LG","authors_text":"Fanjiang Ye, Fengguang Song, Liangliang Cao, Ming Lin, My T. Thai, Tianbao Yang, Xiyuan Wei","submitted_at":"2025-05-10T16:55:03Z","abstract_excerpt":"This paper formalizes an emerging learning paradigm that uses a trained model as a reference to guide and enhance the training of a target model through strategic data selection or weighting, named $\\textbf{model steering}$. While ad-hoc methods have been used in various contexts, including the training of large foundation models, its underlying principles remain insufficiently understood, leading to sub-optimal performance. In this work, we propose a theory-driven framework for model steering called $\\textbf{DRRho risk minimization}$, which is rooted in Distributionally Robust Optimization (D"},"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":"2505.06699","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-10T16:55:03Z","cross_cats_sorted":["cs.AI","cs.CV","stat.ML"],"title_canon_sha256":"0db9f827b2edf4cf56ffe5c1213191f88f0254ea6ad9154b59fec5f4f604d35f","abstract_canon_sha256":"a416a114889cb1585596b99d33af6a651ba9d13f7270210a6c11aa2ce9688b3a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:05:44.067519Z","signature_b64":"F9UCqF5u4wTBWFS/6wA5uoZl200YYFgDBVr9f2jz4pJfra7thY1wlqyTaGnzVU+nutVK4/U4N0eRUfQljYXgCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"25aff2ff37c7dfaf16e37e60ddc50260782d5b1b92dc156bdabbb4c8ea05de55","last_reissued_at":"2026-07-05T11:05:44.067019Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:05:44.067019Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Model Steering: Learning with a Reference Model Improves Generalization Bounds and Scaling Laws","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CV","stat.ML"],"primary_cat":"cs.LG","authors_text":"Fanjiang Ye, Fengguang Song, Liangliang Cao, Ming Lin, My T. Thai, Tianbao Yang, Xiyuan Wei","submitted_at":"2025-05-10T16:55:03Z","abstract_excerpt":"This paper formalizes an emerging learning paradigm that uses a trained model as a reference to guide and enhance the training of a target model through strategic data selection or weighting, named $\\textbf{model steering}$. While ad-hoc methods have been used in various contexts, including the training of large foundation models, its underlying principles remain insufficiently understood, leading to sub-optimal performance. In this work, we propose a theory-driven framework for model steering called $\\textbf{DRRho risk minimization}$, which is rooted in Distributionally Robust Optimization (D"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.06699","kind":"arxiv","version":3},"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/2505.06699/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":"2505.06699","created_at":"2026-07-05T11:05:44.067080+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.06699v3","created_at":"2026-07-05T11:05:44.067080+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.06699","created_at":"2026-07-05T11:05:44.067080+00:00"},{"alias_kind":"pith_short_12","alias_value":"EWX7F7ZXY7P2","created_at":"2026-07-05T11:05:44.067080+00:00"},{"alias_kind":"pith_short_16","alias_value":"EWX7F7ZXY7P26FXD","created_at":"2026-07-05T11:05:44.067080+00:00"},{"alias_kind":"pith_short_8","alias_value":"EWX7F7ZX","created_at":"2026-07-05T11:05:44.067080+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/EWX7F7ZXY7P26FXDPZQN3RICMB","json":"https://pith.science/pith/EWX7F7ZXY7P26FXDPZQN3RICMB.json","graph_json":"https://pith.science/api/pith-number/EWX7F7ZXY7P26FXDPZQN3RICMB/graph.json","events_json":"https://pith.science/api/pith-number/EWX7F7ZXY7P26FXDPZQN3RICMB/events.json","paper":"https://pith.science/paper/EWX7F7ZX"},"agent_actions":{"view_html":"https://pith.science/pith/EWX7F7ZXY7P26FXDPZQN3RICMB","download_json":"https://pith.science/pith/EWX7F7ZXY7P26FXDPZQN3RICMB.json","view_paper":"https://pith.science/paper/EWX7F7ZX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.06699&json=true","fetch_graph":"https://pith.science/api/pith-number/EWX7F7ZXY7P26FXDPZQN3RICMB/graph.json","fetch_events":"https://pith.science/api/pith-number/EWX7F7ZXY7P26FXDPZQN3RICMB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EWX7F7ZXY7P26FXDPZQN3RICMB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EWX7F7ZXY7P26FXDPZQN3RICMB/action/storage_attestation","attest_author":"https://pith.science/pith/EWX7F7ZXY7P26FXDPZQN3RICMB/action/author_attestation","sign_citation":"https://pith.science/pith/EWX7F7ZXY7P26FXDPZQN3RICMB/action/citation_signature","submit_replication":"https://pith.science/pith/EWX7F7ZXY7P26FXDPZQN3RICMB/action/replication_record"}},"created_at":"2026-07-05T11:05:44.067080+00:00","updated_at":"2026-07-05T11:05:44.067080+00:00"}