观察和控制视觉-语言-动作模型中的特征
Observing and Controlling Features in Vision-Language-Action Models
AI总结：
本文提出通过特征可观察性和可控性研究，实现对视觉-语言-动作模型的在线适应与行为引导，提升其与用户需求的实时对齐能力。
摘要：
视觉-语言-动作模型（VLAs）在实现具身智能方面取得了显著进展。尽管其架构在某种程度上与大型语言模型（LLMs）相似，但VLAs由于多模态输入/输出和通常混合的Transformer和扩散头而表现出更高的复杂性。这也是为什么来自LLMs的机械可解释性见解，即解释内部模型表示如何与输出行为相关联，不能轻易转移到VLAs的原因。在本工作中，我们通过引入并分析两个主要概念来缩小这一差距：特征可观察性和特征可控性。特别是，我们首先研究在表示空间中线性编码的特征，并展示如何通过线性分类器来观察它们。然后，我们使用一种基于最优控制的最小线性干预来准确放置内部表示并引导VLAs的输出向期望区域。我们的结果表明，有针对性的、轻量级的干预可以可靠地引导机器人的行为，同时保持闭环能力。我们通过模拟实验在不同的VLAs架构（π_0.5和OpenVLA）上展示了VLAs具有可解释的内部结构，能够在线适应而无需微调，从而实现实时与用户偏好和任务要求的对齐。
英文摘要：
Vision-Language-Action models (VLAs) have shown remarkable progress towards embodied intelligence. While their architecture partially resembles that of Large Language Models (LLMs), VLAs exhibit higher complexity due to their multi-modal inputs/outputs and often hybrid nature of transformer and diffusion heads. This is part of the reason why insights from mechanistic interpretability in LLMs, which explain how the internal model representations relate to their output behavior, do not trivially transfer to VLA counterparts. In this work, we investigate whether VLA internal representations support lightweight behavioral steering without retraining. Across four frontier VLA models, linear \emph{observers} recover state- and action-relevant information in both autoregressive and transformer--flow-matching architectures, and provide robust directions to causally alter VLA outputs. Building on this, we introduce a \emph{controller} that minimally modifies representations to place observer predictions within prescribed target intervals. Closed-loop experiments with $π_{0.5}$ in the LIBERO simulator and on DROID hardware demonstrate improved constraint satisfaction while retaining task performance, with only approximately $1\%$ inference overhead. Together, these experiments show that lightweight linear interventions can reliably steer VLA behavior while preserving closed-loop capabilities, enabling alignment with user preferences and task requirements without fine-tuning.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
