mlsanityData to 7 Oct 2026

GRAFT: Grounded and Efficient Online Reinforcement Adaptation for Fine-Grained Robot Manipulation

arXiv 2608.27079 ↗First posted 27 Aug 2026cs.ROcited by 0

On the map: Vision-language-action models → Reinforcement learning and post-training

Abstract (arXiv)

Pretrained vision-language-action (VLA) policies provide strong priors for robot manipulation, yet adapting them online to fine-grained biomedical tasks remains challenging. Task success often hinges on subtle, view-dependent visual cues, while task-level rewards provide little guidance about which regions matter, making it difficult to learn task-relevant visual grounding from limited real-robot interaction. Online adaptation is further constrained by the computational cost of VLA inference and replay-based updates. We introduce GRAFT (Grounded Reinforcement Adaptation for Fast Task Learning), a framework for efficient online VLA adaptation through grounded perception. GRAFT uses region-level supervision to learn view-specific visual anchors that focus perception on task-relevant local cues without requiring region proposals at deployment. It further combines single-step action generation with cached visual-language prefix reuse to accelerate online learning. Across four biomedical manipulation tasks, GRAFT improves success rates by 32.5 percentage points under matched adaptation budgets, while reducing the computational overhead of online policy updates.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Reinforcement learning and post-training, one of 13 directions in Vision-language-action models. Builds on 36 of its 37 references that are in the map. First posted on arXiv 27 Aug 2026.

Builds on

36 of 37 references.

Closest ancestors · 7

Same direction, newest first

  1. 2026Towards Human-like Physical Intelligence: Lifelong Vision-Language-Action Learning for Robotic Manipulation16 Jul
  2. 2026FORCE: Efficient VLA Reinforcement Fine-Tuning via Value-Calibrated Warm-up and Self-Distillation24 Jun
  3. 2026SOP: A Scalable Online Post-Training System for Vision-Language-Action Models6 Jan
  4. 2025CO-RFT: Efficient Fine-Tuning of Vision-Language-Action Models through Chunked Offline Reinforcement Learning4 Aug
  5. 2025VLA-RL: Towards Masterful and General Robotic Manipulation with Scalable Reinforcement Learning24 May
  6. 2025ConRFT: A Reinforced Fine-tuning Method for VLA Models via Consistency Policy8 Feb
  7. 2025Improving Vision-Language-Action Model with Online Reinforcement Learning28 Jan

Other references · 29

  1. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,777
  2. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,297
  3. 2025Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Successcited by 844
  4. 2025SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Modelcited by 502
  5. 2024TinyVLA: Towards Fast, Data-Efficient Vision-Language-Action Models for Robotic Manipulationcited by 331
  6. 2024TraceVLA: Visual Trace Prompting Enhances Spatial-Temporal Awareness for Generalist Robotic Policiescited by 298
  7. 2024Consistency Policy: Accelerated Visuomotor Policies via Consistency Distillationcited by 159
  8. 2024Precise and Dexterous Robotic Manipulation via Human-in-the-Loop Reinforcement Learningcited by 153
  9. 2025Spatial Forcing: Implicit Spatial Representation Alignment for Vision-language-action Modelcited by 125
  10. 2025PointVLA: Injecting the 3D World into Vision-Language-Action Modelscited by 101
  11. 2025VLA-Cache: Efficient Vision-Language-Action Manipulation via Adaptive Token Cachingcited by 83
  12. 2025ReconVLA: Reconstructive Vision-Language-Action Model as Effective Robot Perceivercited by 80
17 more
  1. 2025RoboGround: Robotic Manipulation with Grounded Vision-Language Priorscited by 50
  2. 2025From Spatial to Actions: Grounding Vision-Language-Action Model in Spatial Foundation Priorscited by 25
  3. 2026ActiveVLA: Injecting Active Perception into Vision-Language-Action Models for Precise 3D Robotic Manipulationcited by 14
  4. 2025AutoBio: A Simulation and Benchmark for Robotic Automation in Digital Biology Laboratorycited by 12
  5. 2026DeltaVLA: Prior-Guided Vision-Language-Action Models via World Knowledge Variationcited by 9
  6. 2024Chemistry3D: Robotic Interaction Benchmark for Chemistry Experimentscited by 8
  7. 2026HiVLA: A Visual-Grounded-Centric Hierarchical Embodied Manipulation Systemcited by 7
  8. 2026GuidedVLA: Specifying Task-Relevant Factors via Plug-and-Play Action Attention Specializationcited by 7
  9. 2025RoboCulture: A Robotics Platform for Automated Biological Experimentationcited by 6
  10. 2025BioMARS: A Multi-Agent Robotic System for Autonomous Biological Experimentscited by 6
  11. 2026PEAfowl: Perception-Enhanced Multi-View Vision-Language-Action for Bimanual Manipulationcited by 6
  12. 2026AffordVLA: Injecting Affordance Representations into Vision-Language-Action Models via Implicit Feature Alignmentcited by 5
  13. 2026BioProVLA-Agent: An Affordable, Protocol-Driven, Vision-Enhanced VLA-Enabled Embodied Multi-Agent System with Closed-Loop-Capable Reasoning for Biological Laboratory Manipulationcited by 3
  14. 2026Pipette: An Embodied Simulation Platform, Benchmark, and Data-Efficient Augmentation Framework for Wet-Lab Roboticscited by 3
  15. 2026LabVLA: Grounding Vision-Language-Action Models in Scientific Laboratoriescited by 3
  16. 2026Never Too Late for Force: Accelerating VLA Post-Training with Reactive Force Injectioncited by 3
  17. 2026Keyframe-Guided Structured Rewards for Reinforcement Learning in Long-Horizon Laboratory Roboticscited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Agentic-VLA: Efficient Online Adaptation for Vision-Language-Action Models
  2. 2026AtomVLA: Scalable Post-Training for Robotic Manipulation via Predictive Latent World Models
  3. 2026PriorVLA: Prior-Preserving Adaptation for Vision-Language-Action Models
  4. 2026RL Token: Bootstrapping Online RL with Vision-Language-Action Models
  5. 2026ROAD-VLA: Robust Online Adaptation via Self-Distillation for Vision-Language-Action Models
  6. 2025EVOLVE-VLA: Test-Time Training from Environment Feedback for Vision-Language-Action Models
  7. 2026RA-VLA: Retrieval-Augmented VLA for Test-Time Adaptation
  8. 2024Learning Efficient and Robust Language-conditioned Manipulation using Textual-Visual Relevancy and Equivariant Language Mapping