{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:7DCTFQTIQLJIZOJFPCJ4RXVWP6","short_pith_number":"pith:7DCTFQTI","schema_version":"1.0","canonical_sha256":"f8c532c26882d28cb9257893c8deb67fa7da814b0e392b9476327a4af1e0698e","source":{"kind":"arxiv","id":"2603.17461","version":2},"attestation_state":"computed","paper":{"title":"AR-CoPO: Align Autoregressive Video Generation with Contrastive Policy Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Bingqi Ma, Dailan He, Guanglu Song, Guanlin Feng, Hongsheng Li, Xingtong Ge, Yi Zhang, Yu Liu","submitted_at":"2026-03-18T08:07:01Z","abstract_excerpt":"Streaming autoregressive (AR) video generators combined with few-step distillation achieve low-latency, high-quality synthesis, yet remain difficult to align via reinforcement learning from human feedback (RLHF). Existing SDE-based GRPO methods face challenges in this setting: few-step ODEs and consistency model samplers deviate from standard flow-matching ODEs, and their short, low-stochasticity trajectories are highly sensitive to initialization noise, rendering intermediate SDE exploration ineffective. We propose AR-CoPO (AutoRegressive Contrastive Policy Optimization), a framework that ada"},"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":"2603.17461","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-03-18T08:07:01Z","cross_cats_sorted":[],"title_canon_sha256":"3025a8620f5bc1928dc1b17dde5009979659cf7a4b6207a2a314049775c699fa","abstract_canon_sha256":"9623378df3e6c92936efba11a2102205bf579a9923626b6feaa312c864813eee"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-01T00:17:10.918828Z","signature_b64":"Ubne14EjOrCvvw/LxfiQnSYBMZTiFUfznXeG5j+34mwm5fIlb6lnxU5H2Zkh2sONPqq0vChRFrBxLh+liFZaDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f8c532c26882d28cb9257893c8deb67fa7da814b0e392b9476327a4af1e0698e","last_reissued_at":"2026-07-01T00:17:10.918329Z","signature_status":"signed_v1","first_computed_at":"2026-07-01T00:17:10.918329Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"AR-CoPO: Align Autoregressive Video Generation with Contrastive Policy Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Bingqi Ma, Dailan He, Guanglu Song, Guanlin Feng, Hongsheng Li, Xingtong Ge, Yi Zhang, Yu Liu","submitted_at":"2026-03-18T08:07:01Z","abstract_excerpt":"Streaming autoregressive (AR) video generators combined with few-step distillation achieve low-latency, high-quality synthesis, yet remain difficult to align via reinforcement learning from human feedback (RLHF). Existing SDE-based GRPO methods face challenges in this setting: few-step ODEs and consistency model samplers deviate from standard flow-matching ODEs, and their short, low-stochasticity trajectories are highly sensitive to initialization noise, rendering intermediate SDE exploration ineffective. We propose AR-CoPO (AutoRegressive Contrastive Policy Optimization), a framework that ada"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.17461","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/2603.17461/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":"2603.17461","created_at":"2026-07-01T00:17:10.918390+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.17461v2","created_at":"2026-07-01T00:17:10.918390+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.17461","created_at":"2026-07-01T00:17:10.918390+00:00"},{"alias_kind":"pith_short_12","alias_value":"7DCTFQTIQLJI","created_at":"2026-07-01T00:17:10.918390+00:00"},{"alias_kind":"pith_short_16","alias_value":"7DCTFQTIQLJIZOJF","created_at":"2026-07-01T00:17:10.918390+00:00"},{"alias_kind":"pith_short_8","alias_value":"7DCTFQTI","created_at":"2026-07-01T00:17:10.918390+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":3,"sample":[{"citing_arxiv_id":"2606.28016","citing_title":"TempAct: Advancing Temporal Plausibility in Autoregressive Video Generation via Planner-Executor RL","ref_index":6,"is_internal_anchor":true},{"citing_arxiv_id":"2606.28016","citing_title":"TempAct: Advancing Temporal Plausibility in Autoregressive Video Generation via Planner-Executor RL","ref_index":6,"is_internal_anchor":true},{"citing_arxiv_id":"2605.14278","citing_title":"KVPO: ODE-Native GRPO for Autoregressive Video Alignment via KV Semantic Exploration","ref_index":2,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7DCTFQTIQLJIZOJFPCJ4RXVWP6","json":"https://pith.science/pith/7DCTFQTIQLJIZOJFPCJ4RXVWP6.json","graph_json":"https://pith.science/api/pith-number/7DCTFQTIQLJIZOJFPCJ4RXVWP6/graph.json","events_json":"https://pith.science/api/pith-number/7DCTFQTIQLJIZOJFPCJ4RXVWP6/events.json","paper":"https://pith.science/paper/7DCTFQTI"},"agent_actions":{"view_html":"https://pith.science/pith/7DCTFQTIQLJIZOJFPCJ4RXVWP6","download_json":"https://pith.science/pith/7DCTFQTIQLJIZOJFPCJ4RXVWP6.json","view_paper":"https://pith.science/paper/7DCTFQTI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.17461&json=true","fetch_graph":"https://pith.science/api/pith-number/7DCTFQTIQLJIZOJFPCJ4RXVWP6/graph.json","fetch_events":"https://pith.science/api/pith-number/7DCTFQTIQLJIZOJFPCJ4RXVWP6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7DCTFQTIQLJIZOJFPCJ4RXVWP6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7DCTFQTIQLJIZOJFPCJ4RXVWP6/action/storage_attestation","attest_author":"https://pith.science/pith/7DCTFQTIQLJIZOJFPCJ4RXVWP6/action/author_attestation","sign_citation":"https://pith.science/pith/7DCTFQTIQLJIZOJFPCJ4RXVWP6/action/citation_signature","submit_replication":"https://pith.science/pith/7DCTFQTIQLJIZOJFPCJ4RXVWP6/action/replication_record"}},"created_at":"2026-07-01T00:17:10.918390+00:00","updated_at":"2026-07-01T00:17:10.918390+00:00"}