FM-Steer：以价值引导的级联去噪增强通用策略
FM-Steer: Enhance Generalist Policies with Value-Guided Cascaded Denoising
对做视觉-语言-动作（VLA）策略的研究者，这篇给出了一个无需重训的测试时增强范式：用中间流验证器对多个噪声动作候选打分并择优，再用轻量去噪器异步精修，从而在保持实时控制频率的同时提升操作性能——可直接迁移到基于扩散/流匹配的视觉运动策略上。
原文依据 a test-time computing framework that augments flow-based Vision-Language-Action (VLA) generalist policies with value-guided sampling and cascaded action denoising
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论人类在物理世界中处理复杂任务时，自然会分配更多时间用于行动前的思考。这一范式近来推动了大语言模型（LLM）在数字域复杂任务上的显著进展。然而，对于与物理世界交互的机器人基础模型而言，测试时计算的潜力仍基本未被探索。在本工作中，我们提出 FM-Steer，一个测试时计算框架，通过价值引导采样与级联动作去噪来增强基于流的视觉-语言-动作（VLA）通用策略，从而为灵巧机器人操作带来更强的控制性能与实时动作频率。FM-Steer 首先引入一个中间流验证器，用于估计候选动作的状态-动作价值。在测试时，策略迭代地采样多个带噪动作提案，并保留预测价值最高的一个，从而无需重新训练即可得到价值对齐的高质量动作。为满足机器人控制严苛的频率要求，FM-Steer 进一步引入级联动作去噪，将开销较大的价值引导采样与快速动作精修解耦。一个轻量级的 Lite-Flow 去噪器异步地接收被选中的高价值带噪动作，并快速去噪以产生最终控制信号，从而实现流畅的高频率执行。在部署期间，中间验证器以低频率运行以提供价值引导采样，而 Lite-Flow 去噪器持续处理被选中的候选动作以维持实时控制。大量实验表明，FM-Steer 在测试时有效地扩展了基于流的 VLA 模型，并在多种仿真基准与真实世界灵巧机器人任务上取得了最优性能。源代码见项目主页：https://hume-vla.github.io。
英文原文摘要 1864 字符
Humans naturally allocate more time before acting when handling complex tasks in the physical world. This paradigm has recently led to remarkable advances in boosting Large Language Models (LLMs) on complex tasks in digital domains. However, the potential of test-time computing remains largely unexplored for robotic foundation models that interact with the physical world. In this work, we propose FM-Steer, a test-time computing framework that augments flow-based Vision-Language-Action (VLA) generalist policies with value-guided sampling and cascaded action denoising, enabling stronger control performance and real-time action rates for dexterous robot manipulation. FM-Steer first introduces an intermediate flow verifier to estimate state-action values for candidate actions. At test time, the policy iteratively samples multiple noisy action proposals and retains the one with the highest predicted value, yielding value-aligned, high-quality actions without retraining. To satisfy the stringent frequency demands of robot control, FM-Steer further introduces cascaded action denoising, decoupling expensive value-guided sampling from fast action refinement. A lightweight Lite-Flow denoiser asynchronously takes the selected high-value noisy action and rapidly denoises it to produce the final control signal, enabling fluid, high-rate execution. During deployment, the intermediate verifier operates at a low frequency to provide value-guided sampling, while the Lite-Flow denoiser continually processes selected candidates to maintain real-time control. Extensive experiments demonstrate that FM-Steer scales flow-based VLA models effectively at test time and achieves state-of-the-art performance across diverse simulation benchmarks and real-world dexterous robotic tasks. The source code is available on the project page: https://hume-vla.github.io.