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

2025-06-21

AI Summary

To address the limited robustness and generalization of vision-language-action (VLA) models in unstructured real-world environments, this paper proposes a test-time scaling framework. It generates diverse action proposals via Gaussian perturbations, constructs an action distribution through majority voting, and employs a vision-language model (VLM)—trained on synthetic data—as an action verifier to select the optimal action. A key contribution is the empirical discovery that action error decays with sample count following an exponential power law, enabling principled co-scaling of sampling and verification. Evaluated on out-of-distribution tasks, the method achieves a +25% absolute performance gain; on in-distribution tasks, it improves by +8%. Further joint fine-tuning with novel robot configurations yields an additional +7% performance boost.

Technology Category

• Computer Vision: Large Vision Models
• Planning, Routing, and Scheduling: Planning with Language Models
• Intelligent Robots: Learning & Optimization for ROB

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

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 8% 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.

Problem

• Enhancing robustness of Vision-Language-Action models in real-world environments
• Scaling test-time sampling and verification for improved model generalization
• Optimizing action selection using synthetic data and VLM-based verification

Innovation

• Test-time scaling with sampling and verification
• Gaussian perturbation and majority voting
• Synthetic data-trained VLM-based verifier