{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:RHOUFHQSA4CLCZF452RYRNQMQV","short_pith_number":"pith:RHOUFHQS","schema_version":"1.0","canonical_sha256":"89dd429e120704b164bceea388b60c857ef7436e73fd41d80ca897575abda10b","source":{"kind":"arxiv","id":"2607.14614","version":1},"attestation_state":"computed","paper":{"title":"Beyond Entropy: Correctness-Aware Advantage Shaping via Contrastive Policy Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Deng Cai, Hao Zhang, Hou Pong Chan, Jia Liu, Long Li, Min Chen, Weiwen Xu","submitted_at":"2026-07-16T06:25:19Z","abstract_excerpt":"Reinforcement learning with verifiable rewards (RLVR) commonly uses entropy for advantage shaping. However, entropy cannot distinguish useful uncertainty from detrimental confusion, limiting its effectiveness as a correctness signal. We propose Contrastive Policy Optimization (CPO), which uses token-level contrastive disagreement between reference-guided and vanilla generation distributions for correctness-aware advantage shaping. Both theoretical and empirical results show that this disagreement reliably indicates token-level correctness. We further show that On-policy Distillation is a speci"},"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":"2607.14614","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-07-16T06:25:19Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"8e671e0939dd929cdc80c159b6f6537dd06efa3f51d1b4c412423a9bce0f89d9","abstract_canon_sha256":"7d1abc49983ac0ea4db77c0927100d89597dba3e7bfe620df96dc46e49fe83a3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-17T01:21:20.942923Z","signature_b64":"ZxAzOJuSuTvjNcXdlaCGdGPYn1KcaaCqyeFgK1ewMMWN8VuU9vnfyPGCbHedi4RhgRU4mdgU2q3o3bm2tpLhDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"89dd429e120704b164bceea388b60c857ef7436e73fd41d80ca897575abda10b","last_reissued_at":"2026-07-17T01:21:20.942077Z","signature_status":"signed_v1","first_computed_at":"2026-07-17T01:21:20.942077Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Beyond Entropy: Correctness-Aware Advantage Shaping via Contrastive Policy Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Deng Cai, Hao Zhang, Hou Pong Chan, Jia Liu, Long Li, Min Chen, Weiwen Xu","submitted_at":"2026-07-16T06:25:19Z","abstract_excerpt":"Reinforcement learning with verifiable rewards (RLVR) commonly uses entropy for advantage shaping. However, entropy cannot distinguish useful uncertainty from detrimental confusion, limiting its effectiveness as a correctness signal. We propose Contrastive Policy Optimization (CPO), which uses token-level contrastive disagreement between reference-guided and vanilla generation distributions for correctness-aware advantage shaping. Both theoretical and empirical results show that this disagreement reliably indicates token-level correctness. We further show that On-policy Distillation is a speci"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.14614","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/2607.14614/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":"2607.14614","created_at":"2026-07-17T01:21:20.942523+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.14614v1","created_at":"2026-07-17T01:21:20.942523+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.14614","created_at":"2026-07-17T01:21:20.942523+00:00"},{"alias_kind":"pith_short_12","alias_value":"RHOUFHQSA4CL","created_at":"2026-07-17T01:21:20.942523+00:00"},{"alias_kind":"pith_short_16","alias_value":"RHOUFHQSA4CLCZF4","created_at":"2026-07-17T01:21:20.942523+00:00"},{"alias_kind":"pith_short_8","alias_value":"RHOUFHQS","created_at":"2026-07-17T01:21:20.942523+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/RHOUFHQSA4CLCZF452RYRNQMQV","json":"https://pith.science/pith/RHOUFHQSA4CLCZF452RYRNQMQV.json","graph_json":"https://pith.science/api/pith-number/RHOUFHQSA4CLCZF452RYRNQMQV/graph.json","events_json":"https://pith.science/api/pith-number/RHOUFHQSA4CLCZF452RYRNQMQV/events.json","paper":"https://pith.science/paper/RHOUFHQS"},"agent_actions":{"view_html":"https://pith.science/pith/RHOUFHQSA4CLCZF452RYRNQMQV","download_json":"https://pith.science/pith/RHOUFHQSA4CLCZF452RYRNQMQV.json","view_paper":"https://pith.science/paper/RHOUFHQS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.14614&json=true","fetch_graph":"https://pith.science/api/pith-number/RHOUFHQSA4CLCZF452RYRNQMQV/graph.json","fetch_events":"https://pith.science/api/pith-number/RHOUFHQSA4CLCZF452RYRNQMQV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RHOUFHQSA4CLCZF452RYRNQMQV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RHOUFHQSA4CLCZF452RYRNQMQV/action/storage_attestation","attest_author":"https://pith.science/pith/RHOUFHQSA4CLCZF452RYRNQMQV/action/author_attestation","sign_citation":"https://pith.science/pith/RHOUFHQSA4CLCZF452RYRNQMQV/action/citation_signature","submit_replication":"https://pith.science/pith/RHOUFHQSA4CLCZF452RYRNQMQV/action/replication_record"}},"created_at":"2026-07-17T01:21:20.942523+00:00","updated_at":"2026-07-17T01:21:20.942523+00:00"}