{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:3LFPIGDENEHMCHGFRUDCXIITTN","short_pith_number":"pith:3LFPIGDE","schema_version":"1.0","canonical_sha256":"dacaf41864690ec11cc58d062ba1139b4b61891524993f358a505d74104e6adb","source":{"kind":"arxiv","id":"2606.11189","version":1},"attestation_state":"computed","paper":{"title":"A Unifying Lens on Supervised Fine-Tuning Through Target Distribution Design","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Cho-Jui Hsieh, Sohyun An, Tong Xie, Yihang Chen, Yuanhao Ban, Yunqi Hong","submitted_at":"2026-06-09T17:59:54Z","abstract_excerpt":"Supervised fine-tuning (SFT) typically maximizes the likelihood of every token in a demonstrated trajectory. However, an observed token can be non-unique, noisy, or misaligned with the model prior. Strictly fitting toward this one-hot target may be suboptimal, especially when the pretrained model encodes a rich knowledge prior. In this work, we reinterpret SFT as target distribution design: instead of studying only the loss objective, we analyze the token-level target that the loss drives the model to match. We introduce the Q-target framework, which decomposes SFT supervision into two explici"},"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":"2606.11189","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-09T17:59:54Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"faed6081f1259efacee05a39565a4929a039a43bd720c09b0acd8bb64b37c138","abstract_canon_sha256":"94d7ec535bac82af12b2a0d2a71e27232d0ff4da8ade963ec78bb944e26489c0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-10T02:12:49.099029Z","signature_b64":"/gfo1ozoOTw57o5Q3q9c3VUqmdhRKAUBlORtXhSXd36BebUat1enPeey75HBLrFXZ3ciPdzsNkXDGxtkVDYhBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"dacaf41864690ec11cc58d062ba1139b4b61891524993f358a505d74104e6adb","last_reissued_at":"2026-06-10T02:12:49.097829Z","signature_status":"signed_v1","first_computed_at":"2026-06-10T02:12:49.097829Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Unifying Lens on Supervised Fine-Tuning Through Target Distribution Design","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Cho-Jui Hsieh, Sohyun An, Tong Xie, Yihang Chen, Yuanhao Ban, Yunqi Hong","submitted_at":"2026-06-09T17:59:54Z","abstract_excerpt":"Supervised fine-tuning (SFT) typically maximizes the likelihood of every token in a demonstrated trajectory. However, an observed token can be non-unique, noisy, or misaligned with the model prior. Strictly fitting toward this one-hot target may be suboptimal, especially when the pretrained model encodes a rich knowledge prior. In this work, we reinterpret SFT as target distribution design: instead of studying only the loss objective, we analyze the token-level target that the loss drives the model to match. We introduce the Q-target framework, which decomposes SFT supervision into two explici"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.11189","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/2606.11189/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":"2606.11189","created_at":"2026-06-10T02:12:49.097983+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.11189v1","created_at":"2026-06-10T02:12:49.097983+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.11189","created_at":"2026-06-10T02:12:49.097983+00:00"},{"alias_kind":"pith_short_12","alias_value":"3LFPIGDENEHM","created_at":"2026-06-10T02:12:49.097983+00:00"},{"alias_kind":"pith_short_16","alias_value":"3LFPIGDENEHMCHGF","created_at":"2026-06-10T02:12:49.097983+00:00"},{"alias_kind":"pith_short_8","alias_value":"3LFPIGDE","created_at":"2026-06-10T02:12:49.097983+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/3LFPIGDENEHMCHGFRUDCXIITTN","json":"https://pith.science/pith/3LFPIGDENEHMCHGFRUDCXIITTN.json","graph_json":"https://pith.science/api/pith-number/3LFPIGDENEHMCHGFRUDCXIITTN/graph.json","events_json":"https://pith.science/api/pith-number/3LFPIGDENEHMCHGFRUDCXIITTN/events.json","paper":"https://pith.science/paper/3LFPIGDE"},"agent_actions":{"view_html":"https://pith.science/pith/3LFPIGDENEHMCHGFRUDCXIITTN","download_json":"https://pith.science/pith/3LFPIGDENEHMCHGFRUDCXIITTN.json","view_paper":"https://pith.science/paper/3LFPIGDE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.11189&json=true","fetch_graph":"https://pith.science/api/pith-number/3LFPIGDENEHMCHGFRUDCXIITTN/graph.json","fetch_events":"https://pith.science/api/pith-number/3LFPIGDENEHMCHGFRUDCXIITTN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3LFPIGDENEHMCHGFRUDCXIITTN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3LFPIGDENEHMCHGFRUDCXIITTN/action/storage_attestation","attest_author":"https://pith.science/pith/3LFPIGDENEHMCHGFRUDCXIITTN/action/author_attestation","sign_citation":"https://pith.science/pith/3LFPIGDENEHMCHGFRUDCXIITTN/action/citation_signature","submit_replication":"https://pith.science/pith/3LFPIGDENEHMCHGFRUDCXIITTN/action/replication_record"}},"created_at":"2026-06-10T02:12:49.097983+00:00","updated_at":"2026-06-10T02:12:49.097983+00:00"}