{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:PCNMERUM4KEKZOQI4VMJ7NJHOS","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"3a1340d177da08657e0d0a1971623f52c995294951d18469f33b5163c4185185","cross_cats_sorted":["cs.RO"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-04T18:36:40Z","title_canon_sha256":"8e90aa034719a618774f63ed25aaa69d7d1dba226830d5d96d7cdf3d53126fe5"},"schema_version":"1.0","source":{"id":"2605.03065","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.03065","created_at":"2026-06-11T01:09:36Z"},{"alias_kind":"arxiv_version","alias_value":"2605.03065v2","created_at":"2026-06-11T01:09:36Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.03065","created_at":"2026-06-11T01:09:36Z"},{"alias_kind":"pith_short_12","alias_value":"PCNMERUM4KEK","created_at":"2026-06-11T01:09:36Z"},{"alias_kind":"pith_short_16","alias_value":"PCNMERUM4KEKZOQI","created_at":"2026-06-11T01:09:36Z"},{"alias_kind":"pith_short_8","alias_value":"PCNMERUM","created_at":"2026-06-11T01:09:36Z"}],"graph_snapshots":[{"event_id":"sha256:b7da2ade31e616663c4dcf27367fda098adafb73a9f5bb05bec9a96d17f3b35e","target":"graph","created_at":"2026-06-11T01:09:36Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"OGPO achieves state-of-the-art performance on manipulation tasks spanning multi-task settings, high-precision insertion, and dexterous control. It is the only method that can fine-tune poorly-initialized behavior cloning policies to near full task-success with no expert data in the online replay buffer, and does so with few task-specific hyperparameter tuning."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That off-policy critics can be maintained stably while propagating gradients through the full generative process of the policy via a modified PPO objective without introducing instability or requiring heavy per-task tuning across state- and pixel-based settings."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"OGPO is a sample-efficient off-policy method for full finetuning of generative control policies that reaches SOTA on robotic manipulation tasks and can recover from poor behavior-cloning initializations without expert data."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"OGPO fine-tunes generative control policies to near full task success from poor initializations with no expert data."}],"snapshot_sha256":"2c21246922662e0d42dda07e7ca12555ba657f22bf3cf7eba0abc5d90f06eb2f"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"4b1db2abbe20e034c83416194efca1c0cfd7d40e207ec3a70c921a05b577b489"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-20T14:38:37.362641Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-20T02:01:22.017696Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T15:44:35.772770Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.03065/integrity.json","findings":[],"snapshot_sha256":"79a51aaa64822ad9c6ff6402b34c61edc00301dc603d44c675650b897c858442","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Generative control policies (GCPs), such as diffusion- and flow-based control policies, have emerged as effective parameterizations for robot learning. This work introduces Off-policy Generative Policy Optimization (OGPO), a sample-efficient algorithm for finetuning GCPs that maintains off-policy critic networks to maximize data reuse and propagate policy gradients through the full generative process of the policy via a modified PPO objective, using critics as the terminal reward. OGPO achieves state-of-the-art performance on manipulation tasks spanning multi-task settings, high-precision inse","authors_text":"Abhishek Gupta, Chaoyi Pan, Cleah Winston, Douglas Chen, Giri Anantharaman, Hongkai Dai, Jesse Zhang, Manan Agarwal, Max Simchowitz, Mitsuhiko Nakamoto, Nai-Chieh Huang, Oliver Kroemer, Paarth Shah, Sarvesh Patil, Sergey Levine, Shashwat Saxena, Zeynep Temel","cross_cats":["cs.RO"],"headline":"OGPO fine-tunes generative control policies to near full task success from poor initializations with no expert data.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-04T18:36:40Z","title":"OGPO: Sample Efficient Full-Finetuning of Generative Control Policies"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.03065","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-08T18:41:53.828658Z","id":"e5507149-99c4-4129-b69a-8b673fed22ce","model_set":{"reader":"grok-4.3"},"one_line_summary":"OGPO is a sample-efficient off-policy method for full finetuning of generative control policies that reaches SOTA on robotic manipulation tasks and can recover from poor behavior-cloning initializations without expert data.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"OGPO fine-tunes generative control policies to near full task success from poor initializations with no expert data.","strongest_claim":"OGPO achieves state-of-the-art performance on manipulation tasks spanning multi-task settings, high-precision insertion, and dexterous control. It is the only method that can fine-tune poorly-initialized behavior cloning policies to near full task-success with no expert data in the online replay buffer, and does so with few task-specific hyperparameter tuning.","weakest_assumption":"That off-policy critics can be maintained stably while propagating gradients through the full generative process of the policy via a modified PPO objective without introducing instability or requiring heavy per-task tuning across state- and pixel-based settings."}},"verdict_id":"e5507149-99c4-4129-b69a-8b673fed22ce"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:1a7859b21ce59f0b2cc3e0afea30e695b8a138995b153557793e1814b346fb7d","target":"record","created_at":"2026-06-11T01:09:36Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"3a1340d177da08657e0d0a1971623f52c995294951d18469f33b5163c4185185","cross_cats_sorted":["cs.RO"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-04T18:36:40Z","title_canon_sha256":"8e90aa034719a618774f63ed25aaa69d7d1dba226830d5d96d7cdf3d53126fe5"},"schema_version":"1.0","source":{"id":"2605.03065","kind":"arxiv","version":2}},"canonical_sha256":"789ac2468ce288acba08e5589fb527749ebe5668847bb47664ccf41ac702ca3a","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"789ac2468ce288acba08e5589fb527749ebe5668847bb47664ccf41ac702ca3a","first_computed_at":"2026-06-11T01:09:36.934198Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-11T01:09:36.934198Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"smqEyff/Vj6HZTYL4/sI02ejykVkACbB7VeAUh0n5dfblccdQIFF7TaDniIkoN8dh+tSl7jvtuz0qbp+/bF7Dw==","signature_status":"signed_v1","signed_at":"2026-06-11T01:09:36.935088Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.03065","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:1a7859b21ce59f0b2cc3e0afea30e695b8a138995b153557793e1814b346fb7d","sha256:b7da2ade31e616663c4dcf27367fda098adafb73a9f5bb05bec9a96d17f3b35e"],"state_sha256":"e88ce164ea9a0398172e34c410a11f3afdf6dfe4c613e4be91a790b1b64ac552"}