TapSampling：基于任务进度理解验证器的推理时采样方法用于机器人操作
TapSampling: Inference-Time Sampling with a Task-Progress-Understanding Verifier for Robotic Manipulation
AI总结：
提出TapSampling框架，通过Action-VAE在低维潜空间采样候选动作，并利用任务进度预测验证器选择最优动作，无需微调即可提升多种通用策略的性能。
摘要：
现有的具身控制研究通过扩展训练数据和模型规模展现了显著的性能提升。我们则探索推理时策略作为另一个维度。非确定性生成模型，如扩散模型和自回归模型，已被广泛应用于具身控制领域。然而，单次推理范式限制了它们的性能。在本文中，我们提出 extbf{TapSampling}，一个即插即用的推理时采样框架。首先，我们引入一个Action-VAE，通过将策略生成的初始动作映射到压缩的后验分布中，在低维潜空间中表示动作，从中可以抽取任意数量的潜样本并解码为候选动作，这些动作近似于真实动作分布。其次，我们将动作验证表述为任务进度结果预测，利用机器人数据集固有的序列结构训练一个语义基础验证器，用于可解释的动作选择。此外，TapSampling是一个策略无关的框架。在模拟和真实环境中的大量实验表明，我们的方法无需进一步微调策略即可显著提升多种通用策略的性能。代码和模型可在项目页面获取。
英文摘要：
Existing embodied control research demonstrates remarkable performance improvements by scaling training data and model size. We instead explore inference-time strategy as an alternative axis. Non-deterministic generative models, such as diffusion and autoregressive models, have been widely adopted in the field of embodied control. However, the single-shot inference paradigm limits their performance. In this paper, we propose \textbf{TapSampling}, a plug-and-play framework for inference-time sampling. First, we introduce an Action-VAE that represents actions in a low-dimensional latent space by mapping policy-generated initial actions into a compressed posterior distribution, from which any number of latent samples can be drawn and decoded into candidate actions that approximate the true action distribution. Second, we formulate action verification as task-progress outcome prediction, using the intrinsic sequential structure of robotic datasets to train a semantically grounded verifier for interpretable action selection. Furthermore, TapSampling is a policy-agnostic framework. Extensive experiments in both simulated and real-world environments demonstrate that our method substantially improves multiple generalist policies without further policy finetuning. Code and models are available at the project page.
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。
