推理作为分布：用于视觉-语言-动作模型的生成式计划隐变量
Reasoning as a Distribution: Generative Plan Latents for Vision-Language-Action Models
AI总结：
该研究针对VLA策略推理压缩为单一表示导致替代方案缺失的问题，提出GPL将推理状态表示为计划分布，实现计划采样、引导与结果适配，在多机器人任务中验证了其引导计划的有效性。
摘要：
视觉-语言-动作（VLA）策略在行动前越来越多地进行推理，但现有推理模型会将中间决策压缩为单一表示。当存在多个有效计划时，这种承诺会隐藏原本可被选择或修改的替代方案。我们提出生成式计划隐变量（GPL），它将推理状态表示为计划的低维条件分布，例如使用哪个对象、抓取方式或路径。这使得推理状态可被处理：无需重新训练VLA，即可采样计划（覆盖性）、按请求选择计划（引导），并根据部署结果重新加权计划（结果适配）。我们在SimplerEnv操纵任务中用π₀测试了这三种功能，并将分析扩展到RoboCasa、LIBERO和迷宫领域。引导计划使遵循请求的执行比例从0.48提升至0.94，而对原生动作分布的相同引导未显示可靠增益。最后，我们表明当显式计划携带观察未确定的决策信息时，其最为有用。
英文摘要：
Vision-Language-Action (VLA) policies increasingly reason before acting, but existing reasoning models compress this intermediate decision into a single representation. When several plans are valid, this commitment hides alternatives that could otherwise be selected or revised. We introduce the Generative Plan Latent (GPL), which represents the reasoning state as a low-dimensional conditional distribution over plans, such as which object, grasp, or route to use. This makes the reasoning state addressable: without retraining the VLA, plans can be sampled (coverage), selected on request (steering), and reweighted by deployment outcomes (outcome adaptation). We test all three on SimplerEnv manipulation with $π_0$, and extend the analysis to RoboCasa, LIBERO, and a maze domain. Steering the plan raises the fraction of executions that follow the request from 0.48 to 0.94, while the same guidance on the native action distribution shows no reliable gain. Finally, we show that an explicit plan is most useful when it carries decision information not already determined by the observation.
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。
