{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:GG7EXLYFYYPGNR6K3W4GJMK5DF","short_pith_number":"pith:GG7EXLYF","schema_version":"1.0","canonical_sha256":"31be4baf05c61e66c7caddb864b15d1973ce72ce53526c62073166150922c10c","source":{"kind":"arxiv","id":"2608.00536","version":1},"attestation_state":"computed","paper":{"title":"DocPO: Advancing Document Policy Optimization via Tailored Step-Aware Rewards","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Binghong Wu, Feng Zhang, Jiacheng Shi, Shuo Huang, Tinghao Yu, Yunhao Wang, Zhenyu Huang","submitted_at":"2026-08-01T08:45:41Z","abstract_excerpt":"Reinforcement learning (RL) for document parsing often relies on reference-based rewards rooted in edit distance (e.g., tree edit distance), yet it remains hard to optimize in the high-accuracy regime because such rewards become weakly discriminative: near-correct outputs receive very similar scores, providing limited learning signal for hard cases. We propose Step-Aware Annealing (SAA), a plug-and-play reward sharpening mechanism that progressively increases reward curvature during training, amplifying subtle quality differences among high-scoring samples while preserving stability in early l"},"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":"2608.00536","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-08-01T08:45:41Z","cross_cats_sorted":[],"title_canon_sha256":"fc869ea23afbf901399bb20c9565ef6479049be3c149739666e87b95e1a6d79e","abstract_canon_sha256":"3127b259872b6d7da069039e871200bb4b053027232f7f22fa911000c35f10b3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-04T00:37:23.765255Z","signature_b64":"L0OnQtT7tIntWO94/YdfOBmuBx7XtKsihj9UVbP1+JTwAuTXFLHpFbjFz2G0KfbeAhuQROUVl5+skv2LxZeQDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"31be4baf05c61e66c7caddb864b15d1973ce72ce53526c62073166150922c10c","last_reissued_at":"2026-08-04T00:37:23.763786Z","signature_status":"signed_v1","first_computed_at":"2026-08-04T00:37:23.763786Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"DocPO: Advancing Document Policy Optimization via Tailored Step-Aware Rewards","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Binghong Wu, Feng Zhang, Jiacheng Shi, Shuo Huang, Tinghao Yu, Yunhao Wang, Zhenyu Huang","submitted_at":"2026-08-01T08:45:41Z","abstract_excerpt":"Reinforcement learning (RL) for document parsing often relies on reference-based rewards rooted in edit distance (e.g., tree edit distance), yet it remains hard to optimize in the high-accuracy regime because such rewards become weakly discriminative: near-correct outputs receive very similar scores, providing limited learning signal for hard cases. We propose Step-Aware Annealing (SAA), a plug-and-play reward sharpening mechanism that progressively increases reward curvature during training, amplifying subtle quality differences among high-scoring samples while preserving stability in early l"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.00536","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/2608.00536/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":"2608.00536","created_at":"2026-08-04T00:37:23.764954+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.00536v1","created_at":"2026-08-04T00:37:23.764954+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.00536","created_at":"2026-08-04T00:37:23.764954+00:00"},{"alias_kind":"pith_short_12","alias_value":"GG7EXLYFYYPG","created_at":"2026-08-04T00:37:23.764954+00:00"},{"alias_kind":"pith_short_16","alias_value":"GG7EXLYFYYPGNR6K","created_at":"2026-08-04T00:37:23.764954+00:00"},{"alias_kind":"pith_short_8","alias_value":"GG7EXLYF","created_at":"2026-08-04T00:37:23.764954+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/GG7EXLYFYYPGNR6K3W4GJMK5DF","json":"https://pith.science/pith/GG7EXLYFYYPGNR6K3W4GJMK5DF.json","graph_json":"https://pith.science/api/pith-number/GG7EXLYFYYPGNR6K3W4GJMK5DF/graph.json","events_json":"https://pith.science/api/pith-number/GG7EXLYFYYPGNR6K3W4GJMK5DF/events.json","paper":"https://pith.science/paper/GG7EXLYF"},"agent_actions":{"view_html":"https://pith.science/pith/GG7EXLYFYYPGNR6K3W4GJMK5DF","download_json":"https://pith.science/pith/GG7EXLYFYYPGNR6K3W4GJMK5DF.json","view_paper":"https://pith.science/paper/GG7EXLYF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.00536&json=true","fetch_graph":"https://pith.science/api/pith-number/GG7EXLYFYYPGNR6K3W4GJMK5DF/graph.json","fetch_events":"https://pith.science/api/pith-number/GG7EXLYFYYPGNR6K3W4GJMK5DF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GG7EXLYFYYPGNR6K3W4GJMK5DF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GG7EXLYFYYPGNR6K3W4GJMK5DF/action/storage_attestation","attest_author":"https://pith.science/pith/GG7EXLYFYYPGNR6K3W4GJMK5DF/action/author_attestation","sign_citation":"https://pith.science/pith/GG7EXLYFYYPGNR6K3W4GJMK5DF/action/citation_signature","submit_replication":"https://pith.science/pith/GG7EXLYFYYPGNR6K3W4GJMK5DF/action/replication_record"}},"created_at":"2026-08-04T00:37:23.764954+00:00","updated_at":"2026-08-04T00:37:23.764954+00:00"}