{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:VZ4ED6DWTNBRTRTTEHZ2K5YPIU","short_pith_number":"pith:VZ4ED6DW","schema_version":"1.0","canonical_sha256":"ae7841f8769b4319c67321f3a5770f4536e15c56af10a6c6f1ad9dd8300599d4","source":{"kind":"arxiv","id":"2502.05450","version":2},"attestation_state":"computed","paper":{"title":"ConRFT: A Reinforced Fine-tuning Method for VLA Models via Consistency Policy","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.RO","authors_text":"Dongbin Zhao, Haoran Li, Shuai Tian, Shugao Liu, Yingting Zhou, Yuhui Chen","submitted_at":"2025-02-08T05:01:17Z","abstract_excerpt":"Vision-Language-Action (VLA) models have shown substantial potential in real-world robotic manipulation. However, fine-tuning these models through supervised learning struggles to achieve robust performance due to limited, inconsistent demonstrations, especially in contact-rich environments. In this paper, we propose a reinforced fine-tuning approach for VLA models, named ConRFT, which consists of offline and online fine-tuning with a unified consistency-based training objective, to address these challenges. In the offline stage, our method integrates behavior cloning and Q-learning to effecti"},"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":"2502.05450","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.RO","submitted_at":"2025-02-08T05:01:17Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"f531d9d4e7d64a63ade669527588cbdc6da82964bcc5cac33e395c17a1b85881","abstract_canon_sha256":"8d3ab5172cdbbb96a655f97e549e294ab3f3b5c6c093a4f1b3a787d04e0b9467"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:48:25.275930Z","signature_b64":"rgce1uQrEgOklJsjQEHUpbRs+C9vpLSllxcu8605n7rgAgK1FS6wgv6ovKzu6X0ovm7gj2UoZHxH1U+Bq3CgCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ae7841f8769b4319c67321f3a5770f4536e15c56af10a6c6f1ad9dd8300599d4","last_reissued_at":"2026-07-05T10:48:25.275400Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:48:25.275400Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ConRFT: A Reinforced Fine-tuning Method for VLA Models via Consistency Policy","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.RO","authors_text":"Dongbin Zhao, Haoran Li, Shuai Tian, Shugao Liu, Yingting Zhou, Yuhui Chen","submitted_at":"2025-02-08T05:01:17Z","abstract_excerpt":"Vision-Language-Action (VLA) models have shown substantial potential in real-world robotic manipulation. However, fine-tuning these models through supervised learning struggles to achieve robust performance due to limited, inconsistent demonstrations, especially in contact-rich environments. In this paper, we propose a reinforced fine-tuning approach for VLA models, named ConRFT, which consists of offline and online fine-tuning with a unified consistency-based training objective, to address these challenges. In the offline stage, our method integrates behavior cloning and Q-learning to effecti"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.05450","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/2502.05450/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":"2502.05450","created_at":"2026-07-05T10:48:25.275460+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.05450v2","created_at":"2026-07-05T10:48:25.275460+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.05450","created_at":"2026-07-05T10:48:25.275460+00:00"},{"alias_kind":"pith_short_12","alias_value":"VZ4ED6DWTNBR","created_at":"2026-07-05T10:48:25.275460+00:00"},{"alias_kind":"pith_short_16","alias_value":"VZ4ED6DWTNBRTRTT","created_at":"2026-07-05T10:48:25.275460+00:00"},{"alias_kind":"pith_short_8","alias_value":"VZ4ED6DW","created_at":"2026-07-05T10:48:25.275460+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":38,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.06706","citing_title":"Vision Language Action (VLA) Models for Unmanned Aerial Robotics and Bimanual Manipulation: A Review","ref_index":94,"is_internal_anchor":true},{"citing_arxiv_id":"2606.26801","citing_title":"Improving Vision-Language-Action Model Fine-Tuning with Structured Stage and Keyframe Supervision","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2606.23640","citing_title":"Learning Process Rewards via Success Visitation Matching for Efficient RL","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2607.01651","citing_title":"One Demonstration Is Enough for Real-World Robotic Reinforcement Learning","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2607.02431","citing_title":"WorldSample: Closed-loop Real-robot RL with World Modelling","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2606.10305","citing_title":"SARM2: Multi-Task Stage Aware Reward Modeling for Self Improving Robotic Manipulation","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2606.10927","citing_title":"AllDayNav: Lifelong Navigation via Real-World Reinforcement Learning","ref_index":58,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09615","citing_title":"DexPIE: Stable Dexterous Policy Improvement from Real-World Experience","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2606.08653","citing_title":"FiberTune: Preserving Action-Fiber Visual Residuals in Vision-Language-Action Fine-Tuning","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2606.03949","citing_title":"Preference-Calibrated Human-in-the-Loop Reinforcement Learning for Robotic Manipulation","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02313","citing_title":"Towards Precise Intent-Aligned VLA Aerial Navigation via Expert-Guided GRPO","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31958","citing_title":"Adapting Generalist Robot Policies with Semantic Reinforcement Learning","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2605.00416","citing_title":"Learning While Deploying: Fleet-Scale Reinforcement Learning for Generalist Robot Policies","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.25477","citing_title":"EXPO-FT: Sample-Efficient Reinforcement Learning Finetuning for Vision-Language-Action Models","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30226","citing_title":"BORA: Bridging Offline Reinforcement Learning and Online Residual Adaptation for Real-World Dexterous VLA Models","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2602.07399","citing_title":"VGAS: Value-Guided Action-Chunk Selection for Few-Shot Vision-Language-Action Adaptation","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2602.10503","citing_title":"Towards Long-Lived Robots: Continual Learning VLA Models via Reinforcement Fine-Tuning","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2602.09023","citing_title":"TwinRL: Digital Twin-Driven Reinforcement Learning for Real-World Robotic Manipulation","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15157","citing_title":"Hand-in-the-Loop: Improving VLA Policies for Dexterous Manipulation via Seamless Hand-Arm Intervention","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17486","citing_title":"DyGRO-VLA: Cross-Task Scaling of Vision-Language-Action Models via Dynamic Grouped Residual Optimization","ref_index":132,"is_internal_anchor":false},{"citing_arxiv_id":"2508.13998","citing_title":"Embodied-R1: Reinforced Embodied Reasoning for General Robotic Manipulation","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2510.12710","citing_title":"Reflection-Based Task Adaptation for Self-Improving VLA","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2511.15669","citing_title":"DeepThinkVLA: Enhancing Reasoning Capability of Vision-Language-Action Models","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2509.08827","citing_title":"A Survey of Reinforcement Learning for Large Reasoning Models","ref_index":77,"is_internal_anchor":false},{"citing_arxiv_id":"2508.13073","citing_title":"Large VLM-based Vision-Language-Action Models for Robotic Manipulation: A Survey","ref_index":187,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VZ4ED6DWTNBRTRTTEHZ2K5YPIU","json":"https://pith.science/pith/VZ4ED6DWTNBRTRTTEHZ2K5YPIU.json","graph_json":"https://pith.science/api/pith-number/VZ4ED6DWTNBRTRTTEHZ2K5YPIU/graph.json","events_json":"https://pith.science/api/pith-number/VZ4ED6DWTNBRTRTTEHZ2K5YPIU/events.json","paper":"https://pith.science/paper/VZ4ED6DW"},"agent_actions":{"view_html":"https://pith.science/pith/VZ4ED6DWTNBRTRTTEHZ2K5YPIU","download_json":"https://pith.science/pith/VZ4ED6DWTNBRTRTTEHZ2K5YPIU.json","view_paper":"https://pith.science/paper/VZ4ED6DW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.05450&json=true","fetch_graph":"https://pith.science/api/pith-number/VZ4ED6DWTNBRTRTTEHZ2K5YPIU/graph.json","fetch_events":"https://pith.science/api/pith-number/VZ4ED6DWTNBRTRTTEHZ2K5YPIU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VZ4ED6DWTNBRTRTTEHZ2K5YPIU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VZ4ED6DWTNBRTRTTEHZ2K5YPIU/action/storage_attestation","attest_author":"https://pith.science/pith/VZ4ED6DWTNBRTRTTEHZ2K5YPIU/action/author_attestation","sign_citation":"https://pith.science/pith/VZ4ED6DWTNBRTRTTEHZ2K5YPIU/action/citation_signature","submit_replication":"https://pith.science/pith/VZ4ED6DWTNBRTRTTEHZ2K5YPIU/action/replication_record"}},"created_at":"2026-07-05T10:48:25.275460+00:00","updated_at":"2026-07-05T10:48:25.275460+00:00"}