mlsanityData to 14 Oct 2026

RoboMonkey: Scaling Test-Time Sampling and Verification for Vision-Language-Action Models

arXiv 2506.17811 ↗First posted 21 Jun 2025cs.RO cs.AI cs.SYcited by 61

On the map: Vision-language-action models → Efficient VLA inference and adaptation

Abstract (arXiv)

Vision-Language-Action (VLA) models have demonstrated remarkable capabilities in visuomotor control, yet ensuring their robustness in unstructured real-world environments remains a persistent challenge. In this paper, we investigate test-time scaling through the lens of sampling and verification as means to enhance the robustness and generalization of VLAs. We first demonstrate that the relationship between action error and the number of generated samples follows an exponentiated power law across a range of VLAs, indicating the existence of inference-time scaling laws. Building on these insights, we introduce RoboMonkey, a test-time scaling framework for VLAs. At deployment, RoboMonkey samples a small set of actions from a VLA, applies Gaussian perturbation and majority voting to construct an action proposal distribution, and then uses a Vision Language Model (VLM)-based verifier to select the optimal action. We propose a synthetic data generation pipeline for training such VLM-based action verifiers, and demonstrate that scaling the synthetic dataset consistently improves verification and downstream accuracy. Through extensive simulated and hardware experiments, we show that pairing existing VLAs with RoboMonkey yields significant performance gains, achieving a 25% absolute improvement on out-of-distribution tasks and 9% on in-distribution tasks. Additionally, when adapting to new robot setups, we show that fine-tuning both VLAs and action verifiers yields a 7% performance increase compared to fine-tuning VLAs alone.

Key facts. Cited by 61 papers in the map (data through 14 Oct 2026). Placed in Efficient VLA inference and adaptation, one of 13 directions in Vision-language-action models. Builds on 45 of its 54 references that are in the map. First posted on arXiv 21 Jun 2025.

Builds on

45 of 54 references.

References · 45

  1. 2017Attention Is All You Needcited by 60,361
  2. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  3. 2022Training language models to follow instructions with human feedbackcited by 15,971
  4. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 13,028
  5. 2021Training Verifiers to Solve Math Word Problemscited by 10,109
  6. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 3,080
  7. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,887
  8. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusioncited by 2,838
  9. 2023PaLM-E: An Embodied Multimodal Language Modelcited by 2,071
  10. 2022RT-1: Robotics Transformer for Real-World Control at Scalecited by 2,061
  11. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameterscited by 1,986
  12. 2023Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardwarecited by 1,568
33 more
  1. 2024Octo: An Open-Source Generalist Robot Policycited by 1,484
  2. 2023LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learningcited by 1,410
  3. 2022Code as Policies: Language Model Programs for Embodied Controlcited by 1,210
  4. 2023SGLang: Efficient Execution of Structured Language Model Programscited by 1,141
  5. 2024Large Language Monkeys: Scaling Inference Compute with Repeated Samplingcited by 991
  6. 2023Open X-Embodiment: Robotic Learning Datasets and RT-X Modelscited by 862
  7. 2024DROID: A Large-Scale In-The-Wild Robot Manipulation Datasetcited by 852
  8. 2023Aligning Large Multimodal Models with Factually Augmented RLHFcited by 712
  9. 2023BridgeData V2: A Dataset for Robot Learning at Scalecited by 657
  10. 2025SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Modelcited by 519
  11. 2024Evaluating Real-World Robot Manipulation Policies in Simulationcited by 495
  12. 2025CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Modelscited by 451
  13. 2024CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulationcited by 422
  14. 2022VIMA: General Robot Manipulation with Multimodal Promptscited by 407
  15. 2024Robotic Control via Embodied Chain-of-Thought Reasoningcited by 346
  16. 2024GR-2: A Generative Video-Language-Action Model with Web-Scale Knowledge for Robot Manipulationcited by 296
  17. 2025Hi Robot: Open-Ended Instruction Following with Hierarchical Vision-Language-Action Modelscited by 221
  18. 2024AlphaMath Almost Zero: Process Supervision without Processcited by 162
  19. 2024The Good, The Bad, and The Greedy: Evaluation of LLMs Should Not Ignore Non-Determinismcited by 116
  20. 2024GRAPE: Generalizing Robot Policy via Preference Alignmentcited by 109
  21. 2024Steering Your Generalists: Improving Robotic Foundation Models via Value Guidancecited by 106
  22. 2025S*: Test Time Scaling for Code Generationcited by 98
  23. 2024Unpacking Failure Modes of Generative Policies: Runtime Monitoring of Consistency and Progresscited by 69
  24. 2024Real-Time Anomaly Detection and Reactive Planning with Large Language Modelscited by 68
  25. 2024Archon: An Architecture Search Framework for Inference-Time Techniquescited by 60
  26. 2025From Foresight to Forethought: VLM-In-the-Loop Policy Steering via Latent Alignmentcited by 52
  27. 2025CodeMonkeys: Scaling Test-Time Compute for Software Engineeringcited by 51
  28. 2024Re-Mix: Optimizing Data Mixtures for Large Scale Imitation Learningcited by 47
  29. 2025How Do Large Language Monkeys Get Their Power (Laws)?cited by 43
  30. 2024Inference-Time Policy Steering through Human Interactionscited by 38
  31. 2022A System-Level View on Out-of-Distribution Data in Roboticscited by 33
  32. 2024Autonomous Improvement of Instruction Following Skills via Foundation Modelscited by 32
  33. 2025Action-Free Reasoning for Policy Generalizationcited by 26

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 10; 2026: 51.

61 papers.

Sort:

By area:

  1. 2026Robometer: Scaling General-Purpose Robotic Reward Models via Trajectory Comparisonscited by 60
  2. 2025Do What You Say: Steering Vision-Language-Action Models via Runtime Reasoning-Action Alignment Verificationcited by 26
  3. 2025NORA-1.5: A Vision-Language-Action Model Trained using World Model- and Action-based Preference Rewardscited by 26
  4. 2025Verifier-free Test-Time Sampling for Vision-Language-Action Modelscited by 23
  5. 2026Scaling Verification Can Be More Effective than Scaling Policy Learning for Vision-Language-Action Alignmentcited by 21
  6. 2026Steerable Vision-Language-Action Policies for Embodied Reasoning and Hierarchical Controlcited by 20
  7. 2025RoVer: Robot Reward Model as Test-Time Verifier for Vision-Language-Action Modelcited by 19
  8. 2026World Action Verifier: Self-Improving World Models via Forward-Inverse Asymmetrycited by 18
  9. 2025Efficient Prediction of Pass@k Scaling in Large Language Modelscited by 16
  10. 2026CycleVLA: Proactive Self-Correcting Vision-Language-Action Models via Subtask Backtracking and Minimum Bayes Risk Decodingcited by 14
  11. 2026On-the-Fly VLA Adaptation via Test-Time Reinforcement Learningcited by 14
  12. 2026Self-Evolving Embodied Agents via Skill-Harness Evolutioncited by 12
38 more
  1. 2026VLA-ATTC: Adaptive Test-Time Compute for VLA Models with Relative Action Critic Modelcited by 9
  2. 2025Towards Deploying VLA without Fine-Tuning: Plug-and-Play Inference-Time VLA Policy Steering via Embodied Evolutionary Diffusioncited by 8
  3. 2025Survey of Vision-Language-Action Models for Embodied Manipulationcited by 7
  4. 2026OmniGuide: Universal Guidance Fields for Enhancing Generalist Robot Policiescited by 6
  5. 2026Neglected Free Lunch from Post-training: Progress Advantage for LLM Agentscited by 6
  6. 2025Pretraining Scaling Laws for Generative Evaluations of Language Modelscited by 5
  7. 2026Action Hallucination in Generative Vision-Language-Action Modelscited by 5
  8. 2026When to Act, Ask, or Learn: Uncertainty-Aware Policy Steeringcited by 5
  9. 2026ReconVLA: An Uncertainty-Guided and Failure-Aware Vision-Language-Action Framework for Robotic Controlcited by 5
  10. 2026Retrieve-then-Steer: Online Success Memory for Test-Time Adaptation of Generative VLAscited by 5
  11. 2026Improving Robotic Generalist Policies via Flow Reversal Steeringcited by 5
  12. 2026τ0-VLA: a Hierarchical Robot Foundation Model with World-Model-Guided Test-Time Computationcited by 5
  13. 2026SCALE: Self-uncertainty Conditioned Adaptive Looking and Execution for Vision-Language-Action Modelscited by 4
  14. 2026Seeing Farther and Smarter: Value-Guided Multi-Path Reflection for VLM Policy Optimizationcited by 4
  15. 2026FASTER: Value-Guided Sampling for Fast RLcited by 4
  16. 2026Look Before You Leap: Distilling Tree Search into Action Evaluation for Frozen VLA Modelscited by 4
  17. 2026EmbodiedSWE: Coding Agents for Long Horizon Dexterous Roboticscited by 4
  18. 2025EVE: A Generator-Verifier System for Generative Policiescited by 3
  19. 2026Self-Supervised Bootstrapping of Action-Predictive Embodied Reasoningcited by 3
  20. 2026Think Twice, Act Once: Verifier-Guided Action Selection For Embodied Agentscited by 3
  21. 2026Q-VGM: Q-Guided Value-Gradient Matching for Offline-to-Online RL of Flow-Matching VLA Policiescited by 3
  22. 2026Visual Verification Enables Inference-time Steering and Autonomous Policy Improvementcited by 3
  23. 2026TEXEDO : Test Time Scaling for Controller-aware Language-conditioned Humanoid Motion Generationcited by 3
  24. 2026ELASTIC: Efficiently Learning to Adaptively Scale Test-Time Compute for Generative Control Policiescited by 3
  25. 2025Evaluating the Robustness of Chinchilla Compute-Optimal Scalingcited by 2
  26. 2026Inference-time Policy Steering via Vision and Touchcited by 2
  27. 2026DREAM-Chunk: Reactive Action Chunking with Latent World Modelcited by 2
  28. 2026Sequential Planning via Anchored Robotic Keypointscited by 2
  29. 2026Geometry Guided Self-Consistency for Physical AIcited by 1
  30. 2026DreamAvoid: Critical-Phase Test-Time Dreaming to Avoid Failures in VLA Policiescited by 1
  31. 2026E-TTS: A New Embodied Test-Time Scaling Framework for Robotic Manipulationcited by 1
  32. 2026Teach and Grow: An Agent-Centered Architecture for General Robot Learningcited by 1
  33. 2026World-Coherent Decoding: Self-Verifying Test-Time Planning for World Action Modelscited by 1
  34. 2026TraceFlow: Guiding Frozen Flow-Matching Robot Policies with Success and Failure Tracescited by 1
  35. 2026In-Context Learning for Robots: Methods and Applicationscited by 1
  36. 2026Mid-Harness: Scaling Actions Between Model and Harness for Terminal Agentscited by 1
  37. 2026Test-Time Perturbation Learning with Delayed Feedback for Vision-Language-Action Modelscited by 0
  38. 2026BOKBO (Best of K Bad Options): Calibrated Abstention for VLA Policiescited by 0

50 most cited of 61 All 61 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025Steering Vision-Language-Action Models as Anti-Exploration: A Test-Time Scaling Approach
  2. 2026Action Draft and Verify: A Self-Verifying Framework for Vision-Language-Action Model
  3. 2026Rethinking Visual-Language-Action Model Scaling: Alignment, Mixture, and Regularization
  4. 2025EdgeVLA: Efficient Vision-Language-Action Models
  5. 2024TinyVLA: Towards Fast, Data-Efficient Vision-Language-Action Models for Robotic Manipulation
  6. 2026ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models
  7. 2026A1: A Fully Transparent Open-Source, Adaptive and Efficient Truncated Vision-Language-Action Model
  8. 2026Rethinking the Practicality of Vision-language-action Model: A Comprehensive Benchmark and An Improved Baseline