# LAW：统一语言、动作与世界建模以实现智能体式物理智能

> 原文链接：https://robotics.hellobike.com/papers/2026-10-16-2610-23334
> 频道：papers | 更新日期：2026-10-16 | 标签：arxiv, cs.RO
> 核心摘要：LAW通过闭环VLM-WAM统一语言推理、动作生成和世界建模，在RoboTwin与长程真机任务上验证。

# LAW：统一语言、动作与世界建模以实现智能体式物理智能

::: v-pre
> 原文链接：[LAW: Unifying Language, Action, and World Modeling for Agentic Physical Intelligence](https://arxiv.org/abs/2610.23334)（arXiv 2610.23334 · 发布日期：2026-10-16）

**LAW: Unifying Language, Action, and World Modeling for Agentic Physical Intelligence**

LAW通过闭环VLM-WAM统一语言推理、动作生成和世界建模，在RoboTwin与长程真机任务上验证。

## 核心要点

- 长程操作需要语言推理来规划子任务目标、评估进展并指导动作选择。
- LAW以语义子任务进展统一VLM与WAM，支持动作融合、状态评估和子任务切换。
- WAM采用Mixture-of-Transformer Experts联合视频-动作建模，并使用双流层级视觉记忆。
- 在50个RoboTwin任务达94.21%；换OpenWAM-alpha达95.56%，并称为开源新SOTA。

**验证范围与局限**：验证范围包括50个RoboTwin任务和三项长程真机任务；真机平均成功率为65%。

## 英文摘要 (Abstract)

World-action models jointly predict actions and future world states, but long-horizon manipulation further requires language reasoning to plan subtask objectives, assess execution progress, and guide action selection. We present LAW, a closed-loop VLM--WAM framework that unifies language reasoning, action generation, and world modeling through semantic subtask progress. Within the LAW agent, a hybrid-expert vision-language model plans the remaining subtasks, estimates the values of and fuses actions proposed by a world-action model, and asynchronously assesses execution status to support continued execution and subtask transitions. Within the world-action model, Mixture-of-Transformer Experts integrates modality-specific and shared computation for joint video--action modeling. Dual-stream hierarchical visual memory maintains temporal context for both models, while subtask-local value learning aligns action evaluation with semantic objectives and preserves supervision from successful prefixes of failed trajectories. LAW achieves 94.21% average success across 50 RoboTwin tasks. Replacing the WAM backbone with OpenWAM-$alpha$ further raises the average success rate to 95.56%, establishing a new open-source state of the art on RoboTwin. Despite having only 2B parameters, our VLM also outperforms variants using Embodied-R1.5 and GPT-5.5 High. On three long-horizon real-robot tasks, LAW achieves 65% average success, compared with 28.33% for $pi_{0.5}$ and 25% for Fast-WAM.

作者：Jiajun Cao, Kai Tang, Bohua Zhang, Zezhong Qian, Ying Li, Xiaozhu Ju, Xiaobao Wei, Jiaming Liu, Yang Ding, Sixiang Chen
:::

## 原始参考来源 (Sources)

- [arXiv 2610.23334：LAW: Unifying Language, Action, and World Modeling for Agentic Physical Intelligence](https://arxiv.org/abs/2610.23334)