论文HF Daily Papers · arXiv 2610.420533 天前

REA运行时实现脑小脑异步协同，Unitree G1任务成功率达95.9%

针对具身系统大模型推理慢与快策略泛化差的矛盾，REA运行时通过路由记忆与边界触发门控实现System 2/1/0异步协同。在LIBERO-Pro测试中成功率达95.9%，控制步数减少25%，并成功迁移至Unitree G1人形机器人真机验证。

AI 整理

本体 SDK 与控制
具身模型

宇树 G1

本文由 AI 根据公开内容整理，可能有误，请以原文为准：REA: A Routed Embodied Agent Runtime for Asynchronous Brain-Cerebellum Control（HF Daily Papers · arXiv 2610.42053）

核心内容

针对基础模型推理耗时高与VLA等快策略泛化能力弱的矛盾，研究提出REA（Routed Embodied Agent）统一运行时架构。该方案确立了“运行中策略不等待大脑”的核心原则，通过解耦 deliberative System 2、embodied router System 1 与 reactive System 0 的异步协作，解决了传统Agent循环阻塞执行或错误回退死锁的问题。在3000次LIBERO-Pro测试中，REA取得95.9%成功率，优于单智能体基线94.7%，同时控制步数与时间分别降低25%和17%。

技术层面，REA采用路由记忆预存策略成功率，利用边界触发裁判门控按需咨询大脑，并设计共享预算升级梯阶实现有界恢复，避免重复触发Fallback。在RoboTwin 2.0测试中，VLA+Code组合成功率达86%，超越MEDex基线。该系统支持学习与编程策略在同一运行时共存，已成功迁移至BEHAVIOR-1K仿真环境及Unitree G1人形机器人真机，在对比方法中取得最高实测成功率，验证了异步脑小脑架构在实体硬件上的有效性。

关键看点

异步协同机制：确立策略不等待大脑原则，利用路由记忆与边界触发门控解耦System 2/1/0，消除Agent循环阻塞与静默失败风险
LIBERO-Pro性能：3000次测试成功率95.9%，超越单智能体基线94.7%，控制步数减少25%，控制耗时降低17%
多策略融合验证：RoboTwin 2.0中VLA+Code成功率86%优于MEDex基线，支持学习策略与程序化策略在同一运行时内无缝协作
真机迁移能力：成功部署于Unitree G1人形机器人及BEHAVIOR-1K环境，在同类对比方法中取得最高实测成功率

工程与产业观察

REA为具身智能提供了标准化的异步脑小脑协同范式，通过路由记忆与预算控制解决大模型落地时的实时性与可靠性瓶颈。其在Unitree G1上的真机验证表明，将高层推理与底层反射解耦是提升人形机器人复杂任务成功率的关键工程路径，对构建通用具身Agent运行时具有重要参考价值。

英文摘要 (Abstract)

Foundation models are becoming remarkably reliable at high-level reasoning, yet they remain too slow and costly to serve as fine-grained controllers. Conversely, fast policies like VLAs and whole-body controllers offer more efficient reflexes, but struggle to generalize. While pairing a deliberative brain (foundation models) with a reactive cerebellum (fast policies) seems natural for embodied systems, two obstacles remain: methods are built and scored on incompatible stacks, and their asynchronous coordination is difficult. For example, an agent that calls the VLA in the agentic loop can stall the fast execution, while a VLA that calls the agent on errors tends to re-trigger fallbacks repeatedly or completely miss silent failures. To address this, we propose REA (Routed Embodied Agent), a unified runtime where a deliberative System 2, an embodied router System 1 and reactive System 0 policies cooperate under a single golden rule - no running policy waits on the brain. Instead of querying the brain at each failure, REA employs a route memory that stores each policy's measured success ahead of time, a boundary-triggered judge gate that consults the brain on policy execution, and a shared-budget escalation ladder for bounded recovery. Evaluated on 3,000 test episodes across 60 LIBERO-Pro tasks, REA reaches 95.9% success and beats the best single-agent baseline in its own policy pool (94.7%) while reducing control steps by 25% and control time by 17%. In a separate 100-episode RoboTwin 2.0 study, REA achieves 84% success with VLA policies and 86% with VLA+code, compared with 80% for the strongest single VLA baseline MEDex. These results demonstrate learned and programmatic policies operating within the same runtime. REA also transfers to BEHAVIOR-1K and a real Unitree G1 humanoid, achieving the highest observed success among the compared methods.

Hugging Face Daily Papers 热度：11 upvotes 作者：Pengfei Zhou, Yixing Ma, Zhengfeiyang Zhang, Haoyuan Chen, Zhenglin Wan, Hexin Wang, Wangbo Zhao, Yang You