论文/arXiv 2610.21915·13 天前

机器人策略应该学习预测什么？重新思考 VLA 学习中的监督目标

论文用带样本级路由的统一 VLA 框架研究监督目标分配，并在 LIBERO、CALVIN 与四个真机任务上验证。

AI 整理

• 具身模型

原文链接：What Should Robot Policies Learn to Predict? Rethinking Supervision Targets in Vision-Language-Action Learning（arXiv 2610.21915 · 发布日期：2026-10-15）

What Should Robot Policies Learn to Predict? Rethinking Supervision Targets in Vision-Language-Action Learning

论文用带样本级路由的统一 VLA 框架研究监督目标分配，并在 LIBERO、CALVIN 与四个真机任务上验证。

核心要点

• VLA 通常以动作预测为主，辅助 LA/FV 目标会与其竞争固定训练预算。
• 方法结合双流多模态扩散 Transformer 与样本级路由，显式分配监督目标。
• 实例化基于 LAPA 及 DINO 变体的 LA 目标，以及三类 FV 表征目标。
• 相同预算下，合适 LA/FV 优于仅动作训练；FV 通常优于 LA。

验证范围与局限：验证范围为 LIBERO、CALVIN 和四个真机任务。

英文摘要 (Abstract)

Vision-language-action (VLA) policies are typically trained with action prediction as the primary supervision signal, since robots ultimately execute actions. Auxiliary objectives such as latent-action (LA) and future-vision (FV) prediction are often added to improve policy learning, yet they compete with action prediction for the same training budget. We ask whether, under a fixed budget, allocating a substantial share of training samples to such objectives, rather than to further action prediction, can improve policy learning. We build a unified VLA framework that couples a two-stream multimodal diffusion transformer (MMDiT) with a sample-level router, which assigns each training sample to either action prediction or an auxiliary objective and thus makes supervision allocation explicit. Within this framework, we instantiate three LA targets based on LAPA and its DINO-feature variants, and three FV targets based on DINOv3, V-JEPA 2, and Wan2.2 representations. Across LIBERO, CALVIN, and four real-robot tasks, suitable LA and FV supervision outperforms action-only training under the same budget, even when action prediction receives no more than half of the training samples. FV generally outperforms LA, although the strongest LA target remains competitive on CALVIN. Diagnostic analyses suggest that DINO-based LA targets emphasize relative motion, whereas FV targets emphasize absolute position and task identity. We therefore argue that what robot policies learn to predict, and how training is allocated among these targets, should be treated as explicit design choices rather than defaulting to action-dominant training.

作者：Haitao Shen, Hanyu Liu, Yihan Lin, Huiping Liu, Junbo Cui, Daoheng Li, Yang Li, Jing Zhang, Wenxin Li, Yuan Yao

来源

本页事实对应的公开页面，以原文为准。

1. arXiv 2610.21915：What Should Robot Policies Learn to Predict? Rethinking Supervision Targets in Vision-Language-Action Learning访问 2026-10-21