{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:ILI2JTPC2LAVCZ6V2MSBGJ7RUV","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":"de19876faf1eddbde8804dc887eb945ac2b1a2b9861aa7505585119aa0f575e2","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.RO","submitted_at":"2025-05-24T14:42:51Z","title_canon_sha256":"b33a2cb43154b37dea226bf00ad80291d2967096372f3629d2434c613d22b641"},"schema_version":"1.0","source":{"id":"2505.18719","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2505.18719","created_at":"2026-05-17T23:38:47Z"},{"alias_kind":"arxiv_version","alias_value":"2505.18719v1","created_at":"2026-05-17T23:38:47Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.18719","created_at":"2026-05-17T23:38:47Z"},{"alias_kind":"pith_short_12","alias_value":"ILI2JTPC2LAV","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"ILI2JTPC2LAVCZ6V","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"ILI2JTPC","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:992c5bde73461760cf369a46f21ae1644aee787f1382efbcf07f0bc3bc080797","target":"graph","created_at":"2026-05-17T23:38:47Z","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":"VLA-RL enables OpenVLA-7B to surpass the strongest finetuned baseline by 4.5% on 40 challenging robotic manipulation tasks in LIBERO, and even matches the performance of advanced commercial models such as π₀-FAST."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That a vision-language model fine-tuned on pseudo reward labels from automatically extracted task segments will provide sufficiently accurate and generalizable rewards for online RL across diverse out-of-distribution scenarios."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"VLA-RL applies online RL to pretrained VLAs, yielding a 4.5% gain over strong baselines on 40 LIBERO manipulation tasks and matching commercial models like π₀-FAST."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"VLA-RL applies online reinforcement learning to raise pretrained vision-language-action models above finetuned baselines on robot tasks."}],"snapshot_sha256":"b6ef84180e0065c11bc639260d7b1bd6818e521518951c230cfdd864f83619d7"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"paper":{"abstract_excerpt":"Recent high-capacity vision-language-action (VLA) models have demonstrated impressive performance on a range of robotic manipulation tasks by imitating human demonstrations. However, exploiting offline data with limited visited states will cause execution failure in out-of-distribution scenarios. Intuitively, an exploration-based method that improves on online collected data at test time could address this limitation. We present VLA-RL, an algorithmic and systematic framework that leverages online reinforcement learning (RL) to improve pretrained auto-regressive VLAs in downstream tasks. Withi","authors_text":"Chubin Zhang, Guanxing Lu, Haonan Jiang, Wenkai Guo, Yansong Tang, Yuheng Zhou, Zifeng Gao, Ziwei Wang","cross_cats":["cs.AI"],"headline":"VLA-RL applies online reinforcement learning to raise pretrained vision-language-action models above finetuned baselines on robot tasks.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.RO","submitted_at":"2025-05-24T14:42:51Z","title":"VLA-RL: Towards Masterful and General Robotic Manipulation with Scalable Reinforcement Learning"},"references":{"count":93,"internal_anchors":36,"resolved_work":93,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Proceedings of Advances in Neural Information Processing Systems (NeurIPS) 35, 28955–28971 (2022) 2, 3","work_id":"2b02057e-69a9-41d7-91b4-09a4ee4adf92","year":2022},{"cited_arxiv_id":"1910.07113","doi":"","is_internal_anchor":true,"ref_index":2,"title":"Solving Rubik's Cube with a Robot Hand","work_id":"81bf9cee-6de8-49f6-b967-3cb853e5ba67","year":1910},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Univg-r1: Reasoning guided universal visual grounding with reinforcement learning","work_id":"fad4943c-f29d-4daa-a7ba-dd6f03aac5aa","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Proceedings of Advances in Neural Information Processing Systems (NeurIPS) 35, 24639– 24654 (2022) 2","work_id":"f271101d-b0ac-4f48-8199-518a4806ce5f","year":2022},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"In: Proceedings of International Conference on Machine Learning (ICML)","work_id":"1973ad0d-d7fd-440f-b9b7-60d76e4519de","year":2023}],"snapshot_sha256":"e731bc05c403e9713eb1696c70072c04f7e7a544ffe099fa2e2283171345a01a"},"source":{"id":"2505.18719","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-16T12:52:28.368258Z","id":"2a57462a-b22a-44d3-90e0-a7abd20e77e6","model_set":{"reader":"grok-4.3"},"one_line_summary":"VLA-RL applies online RL to pretrained VLAs, yielding a 4.5% gain over strong baselines on 40 LIBERO manipulation tasks and matching commercial models like π₀-FAST.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"VLA-RL applies online reinforcement learning to raise pretrained vision-language-action models above finetuned baselines on robot tasks.","strongest_claim":"VLA-RL enables OpenVLA-7B to surpass the strongest finetuned baseline by 4.5% on 40 challenging robotic manipulation tasks in LIBERO, and even matches the performance of advanced commercial models such as π₀-FAST.","weakest_assumption":"That a vision-language model fine-tuned on pseudo reward labels from automatically extracted task segments will provide sufficiently accurate and generalizable rewards for online RL across diverse out-of-distribution scenarios."}},"verdict_id":"2a57462a-b22a-44d3-90e0-a7abd20e77e6"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:1e6cf4a6b8a6c7e5d7ed03d11b852178e300478122d21ed214934148c3fc1517","target":"record","created_at":"2026-05-17T23:38:47Z","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":"de19876faf1eddbde8804dc887eb945ac2b1a2b9861aa7505585119aa0f575e2","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.RO","submitted_at":"2025-05-24T14:42:51Z","title_canon_sha256":"b33a2cb43154b37dea226bf00ad80291d2967096372f3629d2434c613d22b641"},"schema_version":"1.0","source":{"id":"2505.18719","kind":"arxiv","version":1}},"canonical_sha256":"42d1a4cde2d2c15167d5d3241327f1a56bc45dbacb36a1d56519d8c5a32a62f1","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"42d1a4cde2d2c15167d5d3241327f1a56bc45dbacb36a1d56519d8c5a32a62f1","first_computed_at":"2026-05-17T23:38:47.803799Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:38:47.803799Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"OG/rfQxiF+S+hkjnZu/PVdI4YdICApn7LWLm+DPUBoTfVh2UwZjfQ0D1N/jJlB4Nzot06MV9OAC3AM59q7uEAw==","signature_status":"signed_v1","signed_at":"2026-05-17T23:38:47.804302Z","signed_message":"canonical_sha256_bytes"},"source_id":"2505.18719","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:1e6cf4a6b8a6c7e5d7ed03d11b852178e300478122d21ed214934148c3fc1517","sha256:992c5bde73461760cf369a46f21ae1644aee787f1382efbcf07f0bc3bc080797"],"state_sha256":"8ad1e07837b3d000e1514e6de4fd391f8699e94b40aadcaba3ca7475e9ae606f"}