{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:PH55DKONC4VNAXWBEBYMZQJ5R4","short_pith_number":"pith:PH55DKON","schema_version":"1.0","canonical_sha256":"79fbd1a9cd172ad05ec12070ccc13d8f184411bb329fc20a3cd5ea40b7931ea1","source":{"kind":"arxiv","id":"2605.29198","version":1},"attestation_state":"computed","paper":{"title":"Guidance Contrastive Token Credit Assignment for Discrete Policy Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Aditya Grover, Akash Gokul Yusuke Kato, Kazuki Kozuka, Konstantinos Kallidromitis, Shufan Li","submitted_at":"2026-05-28T00:17:49Z","abstract_excerpt":"Group-advantage-based reinforcement learning methods, such as GRPO and DAPO, have demonstrated strong performance across diverse domains, including mathematical reasoning and text-to-image generation. However, their reliance on sample-level rewards introduces a key limitation as uniform credit assignment across all tokens fails to capture fine-grained, token-level contributions. To address this issue, we propose Guidance Contrastive Policy Optimization (GCPO), a novel algorithm that enables per-token credit assignment by contrasting model predictions under positive and negative prompts. Rather"},"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":"2605.29198","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-28T00:17:49Z","cross_cats_sorted":[],"title_canon_sha256":"7dc805be169da93ce613234c25ce3a0c7d8c1119813b0d512707f8eefc99698f","abstract_canon_sha256":"4eaaba20f0c04ad8bc4a3d4caaec402c9573fe75b384aad187e91f8d7482d8de"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-29T01:05:23.879904Z","signature_b64":"WQNJyx1WS2SA5LllbVScvSHUUQyHlSO0masrtkKIFyPwbt1OH0fecsNTuT3R/AGahY+VGc8aNfl4zrc3nN0eDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"79fbd1a9cd172ad05ec12070ccc13d8f184411bb329fc20a3cd5ea40b7931ea1","last_reissued_at":"2026-05-29T01:05:23.879269Z","signature_status":"signed_v1","first_computed_at":"2026-05-29T01:05:23.879269Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Guidance Contrastive Token Credit Assignment for Discrete Policy Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Aditya Grover, Akash Gokul Yusuke Kato, Kazuki Kozuka, Konstantinos Kallidromitis, Shufan Li","submitted_at":"2026-05-28T00:17:49Z","abstract_excerpt":"Group-advantage-based reinforcement learning methods, such as GRPO and DAPO, have demonstrated strong performance across diverse domains, including mathematical reasoning and text-to-image generation. However, their reliance on sample-level rewards introduces a key limitation as uniform credit assignment across all tokens fails to capture fine-grained, token-level contributions. To address this issue, we propose Guidance Contrastive Policy Optimization (GCPO), a novel algorithm that enables per-token credit assignment by contrasting model predictions under positive and negative prompts. Rather"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.29198","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/2605.29198/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":"2605.29198","created_at":"2026-05-29T01:05:23.879394+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.29198v1","created_at":"2026-05-29T01:05:23.879394+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.29198","created_at":"2026-05-29T01:05:23.879394+00:00"},{"alias_kind":"pith_short_12","alias_value":"PH55DKONC4VN","created_at":"2026-05-29T01:05:23.879394+00:00"},{"alias_kind":"pith_short_16","alias_value":"PH55DKONC4VNAXWB","created_at":"2026-05-29T01:05:23.879394+00:00"},{"alias_kind":"pith_short_8","alias_value":"PH55DKON","created_at":"2026-05-29T01:05:23.879394+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/PH55DKONC4VNAXWBEBYMZQJ5R4","json":"https://pith.science/pith/PH55DKONC4VNAXWBEBYMZQJ5R4.json","graph_json":"https://pith.science/api/pith-number/PH55DKONC4VNAXWBEBYMZQJ5R4/graph.json","events_json":"https://pith.science/api/pith-number/PH55DKONC4VNAXWBEBYMZQJ5R4/events.json","paper":"https://pith.science/paper/PH55DKON"},"agent_actions":{"view_html":"https://pith.science/pith/PH55DKONC4VNAXWBEBYMZQJ5R4","download_json":"https://pith.science/pith/PH55DKONC4VNAXWBEBYMZQJ5R4.json","view_paper":"https://pith.science/paper/PH55DKON","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.29198&json=true","fetch_graph":"https://pith.science/api/pith-number/PH55DKONC4VNAXWBEBYMZQJ5R4/graph.json","fetch_events":"https://pith.science/api/pith-number/PH55DKONC4VNAXWBEBYMZQJ5R4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PH55DKONC4VNAXWBEBYMZQJ5R4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PH55DKONC4VNAXWBEBYMZQJ5R4/action/storage_attestation","attest_author":"https://pith.science/pith/PH55DKONC4VNAXWBEBYMZQJ5R4/action/author_attestation","sign_citation":"https://pith.science/pith/PH55DKONC4VNAXWBEBYMZQJ5R4/action/citation_signature","submit_replication":"https://pith.science/pith/PH55DKONC4VNAXWBEBYMZQJ5R4/action/replication_record"}},"created_at":"2026-05-29T01:05:23.879394+00:00","updated_at":"2026-05-29T01:05:23.879394+00:00"}