论文/arXiv 2610.12194·20 天前

MiniWAM：为高效 World-Action Modeling 学习紧凑未来目标

针对 WAM 未来状态目标高维且训练成本高的问题，MiniWAM 用 PRISM 学习紧凑未来表示，并在多个仿真基准验证。

AI 整理

• 本体 SDK 与控制
• 仿真与训练
• 具身模型

原文链接：MiniWAM: Learning Compact Future Targets for Efficient World-Action Modeling（arXiv 2610.12194 · 发布日期：2026-10-08）

MiniWAM: Learning Compact Future Targets for Efficient World-Action Modeling

针对 WAM 未来状态目标高维且训练成本高的问题，MiniWAM 用 PRISM 学习紧凑未来表示，并在多个仿真基准验证。

核心要点

• 现有 WAM 多在预训练视觉骨干的原生表示空间预测未来状态，目标高维、训练成本较高。
• MiniWAM 通过 PRISM 从特权当前-未来转移中学习紧凑目标，再联合预测该目标与机器人动作。
• 在 DINOv3 和 WAN2.1 VAE 特征下，MiniWAM 优于原生未来特征预测，并最高加速 8 倍训练。
• 0.25B 参数的 MiniWAM 在 LIBERO、LIBERO-Plus 和 RoboTwin 2.0 仿真基准上已具竞争力，项目页已公开。

验证范围与局限：验证范围为 LIBERO、LIBERO-Plus 和 RoboTwin 2.0 仿真基准。

英文摘要 (Abstract)

World modeling has emerged as an effective co-training objective for robot policies, giving rise to World Action Models (WAMs) that jointly predict actions and future states. However, most WAMs predict future states in the native representation space of pretrained visual backbones, resulting in high-dimensional targets with substantial training cost. We introduce MiniWAM, which instead predicts compact future representations learned from privileged current-future transitions. To construct these targets, we propose Predictive Representations via Inverse Spatiotemporal Modeling (PRISM), which combines inverse-dynamics supervision with feature reconstruction to emphasize control-relevant transition information while preserving useful future-state information. With the learned PRISM encoder frozen, MiniWAM is trained to jointly predict the resulting targets and robot actions from current observations. With 65$\times$ fewer native future feature tokens, MiniWAM consistently outperforms native future-feature prediction with both DINOv3 and WAN2.1 VAE features, while achieving up to an 8$\times$ speedup in world-action training. At 0.25B parameters, MiniWAM is already competitive with substantially larger WAMs on LIBERO, LIBERO-Plus, and RoboTwin 2.0 simulation benchmarks. Representation analyses further show that PRISM contributes behavioral structure beyond feature reconstruction alone. These results demonstrate that effective world-action modeling does not require predicting native visual futures, and that compact predictive representations provide a strong and substantially more efficient target for policy learning. The project page is available at: https://j1dan.github.io/MiniWAM.

作者：Jie Chen, Ruofei Bai, Yuxin Cai, Yifeng Zhang, Chengyang He, Jun Li, Wei Yun Yau, Guillaume Sartoretti

来源

本页事实对应的公开页面，以原文为准。

1. arXiv 2610.12194：MiniWAM: Learning Compact Future Targets for Efficient World-Action Modeling访问 2026-10-21