AI 中文总结
本研究针对视觉-语言-动作（VLA）模型的本体感受状态接入方式开展受控实验，探究状态接入位置、历史长度等问题，得出系统性答案并提炼为可验证的VLA模型设计原则。
AI 中文摘要
近期的视觉-语言-动作（Vision-Language-Action, VLA）模型几乎普遍将机器人本体感受状态作为输入，但采用的接入方式互不兼容——或序列化为文本提示词，或投影到视觉-语言前缀，或直接输入动作专家模块，且几乎均以单个当前帧的形式提供。目前仍存在三个未解决的问题：（1）当前状态是否能改善闭环控制，以及在哪些任务上能改善；（2）状态历史的帮助程度如何，其益处是否反映了真实的时间变化而非额外的条件容量；（3）状态应接入模型的哪个部分——视觉-语言主干还是动作生成模块。我们通过对基于流匹配（flow-matching）的VLA模型开展受控实验来回答这些问题，实验全程固定主干、训练数据、动作表示及评估协议。我们在匹配的实现细节下实现了五种代表性接入方式——离散状态提示词、VLM前缀、动作前缀、状态专家、特征调制，并在涵盖三类任务家族的45个原子任务及20个复合任务上对其进行评估；随后将状态历史长度从1帧扩展至96帧，以探究历史状态信息如何影响模型性能。实验得出了对所有三个问题的系统性答案，提炼为可验证的感知状态感知VLA模型的设计原则。
英文摘要
Recent Vision-Language-Action (VLA) models almost universally take robot proprioceptive state as input, yet wire it in incompatible ways -- serialized into text prompts, projected into the vision-language prefix, or fed directly to the action expert -- and almost always as a single current frame. Three questions remain open: (1) whether, and on which tasks, current state actually improves closed-loop control; (2) how much state history helps, and whether its benefit reflects genuine temporal variation rather than added conditioning capacity; and (3) where state should enter the model -- the vision-language backbone or the action-generation module. We answer these questions through controlled experiments on a flow-matching VLA, fixing the backbone, training data, action representation, and evaluation protocol throughout. We implement five representative interfaces -- discrete state prompt, VLM prefix, action prefix, state expert, and feature modulation -- under matched implementation details, and evaluate them on 45 atomic tasks spanning three task families plus 20 composite tasks; we then sweep the state-history length from 1 to 96 frames to examine how historical state information affects model performance. The experiments yield systematic answers to all three questions, distilled into testable design principles for state-aware VLAs.
