先学移动，再学执行：面向VLA的任务无关预训练

摘要

视觉-语言-动作（VLA）模型的核心瓶颈在于专家示范数据的稀缺——即观测、指令和动作的三元组，这类数据在大规模采集时成本高昂。我们认为，这一瓶颈源于将两个不同的学习目标混为一谈：获取物理能力（如何移动）和获取语义对齐（做什么）。关键的是，只有后者需要语言监督。基于这一分解假设，我们提出了任务无关预训练（TAP）——一种两阶段框架，首先通过自监督的逆动力学目标，从廉价的无标签交互数据（包括离任务轨迹和自主机器人操作）中学习可迁移的运动先验；第二阶段则利用少量专家数据，将这些先验与语言进行轻量级的对齐。在SIMPLER基准上，TAP在使用的标注数据量少数个数量级的情况下，达到了与基于超过100万条专家轨迹训练的模型相当的性能，相比标准行为克隆方法取得了10%的绝对提升。在真实世界的WidowX平台上，TAP在相机扰动下仍保留了25%的成功率，而依赖互联网规模数据训练的基线模型则完全崩溃（0%成功率）。这表明，任务无关预训练能够产生鲁棒、可迁移的物理表征，为具身人工智能提供了一条可扩展的前进路径。

English

Vision-Language-Action (VLA) models are fundamentally bottlenecked by the scarcity of expert demonstrations -- triplets of observations, instructions, and actions that are costly to collect at scale. We argue that this bottleneck stems from conflating two distinct learning objectives: acquiring physical competence (how to move) and acquiring semantic alignment (what to do). Crucially, only the latter requires language supervision. Building on this Decomposition Hypothesis, we propose Task-Agnostic Pretraining (TAP), a two-stage framework that first learns transferable motor priors from cheap, unlabeled interaction data -- including discarded off-task trajectories and autonomous robot play -- via a self-supervised Inverse Dynamics objective. A lightweight second stage then grounds these priors in language using minimal expert data. On the SIMPLER benchmark, TAP matches models trained on over 1M expert trajectories while using orders of magnitude less labeled data, yielding a 10% absolute gain over standard behavior cloning. On a real-world WidowX platform, TAP retains 25% success under camera perturbations where internet-scale baselines collapse to 0%, demonstrating that task-agnostic pretraining produces robust, transferable physical representations and offers a scalable path forward for Embodied AI.