{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:EG7C5N4DPGZXYCAFS4NARYWIYF","short_pith_number":"pith:EG7C5N4D","schema_version":"1.0","canonical_sha256":"21be2eb78379b37c0805971a08e2c8c15635c41199d97de8ec660a25b18ffd92","source":{"kind":"arxiv","id":"2603.00025","version":2},"attestation_state":"computed","paper":{"title":"TAB-PO: Preference Optimization with a Token-Level Adaptive Barrier for Token-Critical Structured Generation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Afshan Khan, Aimee Kendall Roundtree, Ashley Hagaman, Elyas Irankhah, Ganesh Puthiaraju, Linhai Ma, Muhammad Arif, Samah Fodeh, Sarah R. Lowe, Sreeraj Ramachandran, Srivani Talakokkul","submitted_at":"2026-02-03T23:48:13Z","abstract_excerpt":"Direct Preference Optimization (DPO) is an effective and widely adopted approach for offline alignment but is poorly matched to ontology-driven structured prediction, where preferred and rejected JSON objects often differ in only a few schema-defining tokens. In this low-edit-distance regime, sequence-level DPO spreads gradient mass across non-critical serialization tokens (gradient dilution) and can reduce likelihood on rare, under-confident preferred schema tokens (token erosion). To address these limitations, we first develop a confusion-aware preference-construction strategy that augments "},"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":"2603.00025","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-02-03T23:48:13Z","cross_cats_sorted":[],"title_canon_sha256":"7c5fd2bbb7be7fd4d4fa6ef79cf8e4d58e0af9ddf2e8bd668ec79b9b253f7576","abstract_canon_sha256":"dd618cdbd041119f8728b1c2e72eb128ff21ba9910c98a6c9a622646f599b096"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-12T01:08:23.376010Z","signature_b64":"/T4UTBK2CGaq+ceRKZ+b8Me4AJEjxN/ZC+TlnK0TIkDUiKCBHk/iXUW+J/new2VfAuGXZ3+j0CAHIVCXgU9dCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"21be2eb78379b37c0805971a08e2c8c15635c41199d97de8ec660a25b18ffd92","last_reissued_at":"2026-06-12T01:08:23.374927Z","signature_status":"signed_v1","first_computed_at":"2026-06-12T01:08:23.374927Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"TAB-PO: Preference Optimization with a Token-Level Adaptive Barrier for Token-Critical Structured Generation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Afshan Khan, Aimee Kendall Roundtree, Ashley Hagaman, Elyas Irankhah, Ganesh Puthiaraju, Linhai Ma, Muhammad Arif, Samah Fodeh, Sarah R. Lowe, Sreeraj Ramachandran, Srivani Talakokkul","submitted_at":"2026-02-03T23:48:13Z","abstract_excerpt":"Direct Preference Optimization (DPO) is an effective and widely adopted approach for offline alignment but is poorly matched to ontology-driven structured prediction, where preferred and rejected JSON objects often differ in only a few schema-defining tokens. In this low-edit-distance regime, sequence-level DPO spreads gradient mass across non-critical serialization tokens (gradient dilution) and can reduce likelihood on rare, under-confident preferred schema tokens (token erosion). To address these limitations, we first develop a confusion-aware preference-construction strategy that augments "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.00025","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/2603.00025/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":"2603.00025","created_at":"2026-06-12T01:08:23.375073+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.00025v2","created_at":"2026-06-12T01:08:23.375073+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.00025","created_at":"2026-06-12T01:08:23.375073+00:00"},{"alias_kind":"pith_short_12","alias_value":"EG7C5N4DPGZX","created_at":"2026-06-12T01:08:23.375073+00:00"},{"alias_kind":"pith_short_16","alias_value":"EG7C5N4DPGZXYCAF","created_at":"2026-06-12T01:08:23.375073+00:00"},{"alias_kind":"pith_short_8","alias_value":"EG7C5N4D","created_at":"2026-06-12T01:08:23.375073+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":3,"sample":[{"citing_arxiv_id":"2605.24172","citing_title":"EPPC-OASIS: Ontology-Aware Adaptation and Structured Inference Refinement for Electronic Patient-Provider Communication Mining in Secure Messages","ref_index":13,"is_internal_anchor":true},{"citing_arxiv_id":"2604.09737","citing_title":"STaR-DRO: Stateful Tsallis Reweighting for Group-Robust Structured Prediction","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2604.07116","citing_title":"Yale-DM-Lab at ArchEHR-QA 2026: Deterministic Grounding and Multi-Pass Evidence Alignment for EHR Question Answering","ref_index":10,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/EG7C5N4DPGZXYCAFS4NARYWIYF","json":"https://pith.science/pith/EG7C5N4DPGZXYCAFS4NARYWIYF.json","graph_json":"https://pith.science/api/pith-number/EG7C5N4DPGZXYCAFS4NARYWIYF/graph.json","events_json":"https://pith.science/api/pith-number/EG7C5N4DPGZXYCAFS4NARYWIYF/events.json","paper":"https://pith.science/paper/EG7C5N4D"},"agent_actions":{"view_html":"https://pith.science/pith/EG7C5N4DPGZXYCAFS4NARYWIYF","download_json":"https://pith.science/pith/EG7C5N4DPGZXYCAFS4NARYWIYF.json","view_paper":"https://pith.science/paper/EG7C5N4D","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.00025&json=true","fetch_graph":"https://pith.science/api/pith-number/EG7C5N4DPGZXYCAFS4NARYWIYF/graph.json","fetch_events":"https://pith.science/api/pith-number/EG7C5N4DPGZXYCAFS4NARYWIYF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EG7C5N4DPGZXYCAFS4NARYWIYF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EG7C5N4DPGZXYCAFS4NARYWIYF/action/storage_attestation","attest_author":"https://pith.science/pith/EG7C5N4DPGZXYCAFS4NARYWIYF/action/author_attestation","sign_citation":"https://pith.science/pith/EG7C5N4DPGZXYCAFS4NARYWIYF/action/citation_signature","submit_replication":"https://pith.science/pith/EG7C5N4DPGZXYCAFS4NARYWIYF/action/replication_record"}},"created_at":"2026-06-12T01:08:23.375073+00:00","updated_at":"2026-06-12T01:08:23.375073+00:00"}