# 双臂视觉-语言-动作模型中的臂级组合泛化

> 原文链接：https://robotics.hellobike.com/papers/2026-10-05-2610-06184
> 频道：papers | 更新日期：2026-10-05 | 标签：arxiv, cs.RO
> 核心摘要：提出 ACG-Bench 并基于 π0.5 比较数据增强与架构策略，在仿真和 SO101 真机验证双臂技能重组泛化。

# 双臂视觉-语言-动作模型中的臂级组合泛化

::: v-pre
> 原文链接：[Arm-wise Compositional Generalization in Dual-Arm Vision-Language-Action Models](https://arxiv.org/abs/2610.06184)（arXiv 2610.06184 · 发布日期：2026-10-05）

**Arm-wise Compositional Generalization in Dual-Arm Vision-Language-Action Models**

提出 ACG-Bench 并基于 π0.5 比较数据增强与架构策略，在仿真和 SO101 真机验证双臂技能重组泛化。

## 核心要点

- 现有评测难以揭示不同协调需求下支撑跨臂技能重组的训练和架构选择。
- ACG-Bench提供双臂策略测试床，包含23个任务-条件对和多类未见组合。
- 研究臂 token 分组、SkillLoRA 与 AWA，显示技能条件参数和注意力结构互补。
- AE-VLA在仿真和SO101真机上均高于所比较基线，给出双臂策略设计经验。

**验证范围与局限**：验证范围限于仿真中的23个任务-条件对及SO101真机上的五个未见条件。

## 英文摘要 (Abstract)

Generalization in multi-arm collaboration can be studied as composing familiar atomic skills in new ways across arms. However, existing evaluations offer limited insight into which training and architectural choices support this ability under different coordination requirements. We introduce \textbf{ACG-Bench}, a benchmark for \emph{Arm-wise Compositional Generalization} that provides a common testbed for studying skill recomposition in dual-arm policies. It contains 23 task--condition pairs across 8 task families, with 6 in-domain conditions and 17 unseen compositions covering reordering, synchronization, their combination, and cross-task composition. All methods receive the same per-arm atomic prompts, and success requires achieving the task goal while satisfying physical milestones and specified order or timing constraints. Using $π_{0.5}$ as a common vision-language-action backbone, we compare representative data-augmentation and architectural strategies with shared source data and a common evaluation protocol. Our architectural study examines arm-token grouping, skill-specific LoRA adapters (SkillLoRA), and arm-wise attention (AWA), highlighting the complementarity of skill-conditioned parameters and attention structure. Combining these choices yields \textbf{AE-VLA}, which achieves 21.53\% generalization success in simulation, compared with 2.94\% for Single $π_{0.5}$, 3.06\% for MA-VLA, and 5.53\% for two independently controlled $π_{0.5}$ policies. On physical SO101 robots, AE-VLA reaches 39.00\% mean success across five unseen conditions, compared with 10.00\% for the strongest baseline. These findings provide empirical guidance for designing dual-arm policies that generalize beyond fixed training routines.

作者：Zaibin Zhang, Binghao Ran, Yuhan Wu, Zhongbo Zhang, Yifan Wang, Junwei Jiang, Junlan Xiao, Shi, Wangcheng, Li Kang, Yiran Qin
:::

## 原始参考来源 (Sources)

- [arXiv 2610.06184：Arm-wise Compositional Generalization in Dual-Arm Vision-Language-Action Models](https://arxiv.org/abs/2610.06184)