{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:UWFYZCPYWIV37QXH6SVRVGJXT6","short_pith_number":"pith:UWFYZCPY","schema_version":"1.0","canonical_sha256":"a58b8c89f8b22bbfc2e7f4ab1a99379f96b013b5c84c0218abf14a0df70b0e10","source":{"kind":"arxiv","id":"2605.21883","version":1},"attestation_state":"computed","paper":{"title":"Token-weighted Direct Preference Optimization with Attention","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chengyu Huang, Claire Cardie, Sheng-Yen Chou, Zhuohang Li","submitted_at":"2026-05-21T01:43:09Z","abstract_excerpt":"Direct Preference Optimization (DPO) aligns Large Language Models with human preferences without the need for a separate reward model. However, DPO treats all tokens in responses equally, neglecting the differing importance of individual tokens. Existing token-level PO methods compute the token weights using either token-position-based heuristic functions or probability estimates given by a separately trained model, which lacks robustness and incurs extra training cost. In contrast, we propose Token-weighted DPO (TwDPO) -- a novel training objective grounded on token-weighted RL -- and Attenti"},"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":"2605.21883","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-21T01:43:09Z","cross_cats_sorted":[],"title_canon_sha256":"f9a38e902b4444608d0f04ca40ffee80305fa03b830992c3bfd1f9c3b1e53517","abstract_canon_sha256":"c18576ffca0c9b14870d93fa386422af1fd2be1d02a393d8382b34d4d69fc7ec"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-22T01:04:12.975982Z","signature_b64":"Gc1Vyo3TaNWvxdNjDC5qbKQOlaNAtaLiKYGWUVCJ+RK/JbfxD+y/vsaPpZFXlWsZmAN3dFYpsX3dUXMFsr0XCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a58b8c89f8b22bbfc2e7f4ab1a99379f96b013b5c84c0218abf14a0df70b0e10","last_reissued_at":"2026-05-22T01:04:12.975190Z","signature_status":"signed_v1","first_computed_at":"2026-05-22T01:04:12.975190Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Token-weighted Direct Preference Optimization with Attention","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chengyu Huang, Claire Cardie, Sheng-Yen Chou, Zhuohang Li","submitted_at":"2026-05-21T01:43:09Z","abstract_excerpt":"Direct Preference Optimization (DPO) aligns Large Language Models with human preferences without the need for a separate reward model. However, DPO treats all tokens in responses equally, neglecting the differing importance of individual tokens. Existing token-level PO methods compute the token weights using either token-position-based heuristic functions or probability estimates given by a separately trained model, which lacks robustness and incurs extra training cost. In contrast, we propose Token-weighted DPO (TwDPO) -- a novel training objective grounded on token-weighted RL -- and Attenti"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.21883","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/2605.21883/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":"2605.21883","created_at":"2026-05-22T01:04:12.975336+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.21883v1","created_at":"2026-05-22T01:04:12.975336+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.21883","created_at":"2026-05-22T01:04:12.975336+00:00"},{"alias_kind":"pith_short_12","alias_value":"UWFYZCPYWIV3","created_at":"2026-05-22T01:04:12.975336+00:00"},{"alias_kind":"pith_short_16","alias_value":"UWFYZCPYWIV37QXH","created_at":"2026-05-22T01:04:12.975336+00:00"},{"alias_kind":"pith_short_8","alias_value":"UWFYZCPY","created_at":"2026-05-22T01:04:12.975336+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/UWFYZCPYWIV37QXH6SVRVGJXT6","json":"https://pith.science/pith/UWFYZCPYWIV37QXH6SVRVGJXT6.json","graph_json":"https://pith.science/api/pith-number/UWFYZCPYWIV37QXH6SVRVGJXT6/graph.json","events_json":"https://pith.science/api/pith-number/UWFYZCPYWIV37QXH6SVRVGJXT6/events.json","paper":"https://pith.science/paper/UWFYZCPY"},"agent_actions":{"view_html":"https://pith.science/pith/UWFYZCPYWIV37QXH6SVRVGJXT6","download_json":"https://pith.science/pith/UWFYZCPYWIV37QXH6SVRVGJXT6.json","view_paper":"https://pith.science/paper/UWFYZCPY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.21883&json=true","fetch_graph":"https://pith.science/api/pith-number/UWFYZCPYWIV37QXH6SVRVGJXT6/graph.json","fetch_events":"https://pith.science/api/pith-number/UWFYZCPYWIV37QXH6SVRVGJXT6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UWFYZCPYWIV37QXH6SVRVGJXT6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UWFYZCPYWIV37QXH6SVRVGJXT6/action/storage_attestation","attest_author":"https://pith.science/pith/UWFYZCPYWIV37QXH6SVRVGJXT6/action/author_attestation","sign_citation":"https://pith.science/pith/UWFYZCPYWIV37QXH6SVRVGJXT6/action/citation_signature","submit_replication":"https://pith.science/pith/UWFYZCPYWIV37QXH6SVRVGJXT6/action/replication_record"}},"created_at":"2026-05-22T01:04:12.975336+00:00","updated_at":"2026-05-22T01:04:12.975336+00:00"}