# StairVLA：面向 Vision-Language-Action 模型的阶段感知分层动作生成

> 原文链接：https://robotics.hellobike.com/papers/2026-10-06-2610-07756
> 频道：papers | 更新日期：2026-10-06 | 标签：arxiv, cs.RO
> 核心摘要：StairVLA针对VLA动作头均匀去噪的问题，以阶段感知分层生成在仿真基准和真机任务上验证降本保性能。

# StairVLA：面向 Vision-Language-Action 模型的阶段感知分层动作生成

::: v-pre
> 原文链接：[StairVLA: Stage-Aware Hierarchical Action Generation for Vision-Language-Action Models](https://arxiv.org/abs/2610.07756)（arXiv 2610.07756 · 发布日期：2026-10-06）

**StairVLA: Stage-Aware Hierarchical Action Generation for Vision-Language-Action Models**

StairVLA针对VLA动作头均匀去噪的问题，以阶段感知分层生成在仿真基准和真机任务上验证降本保性能。

## 核心要点

- 动机是扩散或流匹配动作头常均匀处理去噪轨迹，但不同阶段的条件关注会自然变化。
- 方法用部分去噪动作连接长时程粗生成与局部细化，高层VLA早期去噪，轻量细化器高频修正。
- 在LIBERO上，GR00T风格实例平均成功率由96.5%升至97.8%，每动作块延迟由115.0 ms降至44.2 ms。
- 跨两个VLA骨干、仿真基准和真机任务，StairVLA持续降低推理成本并保持较强任务表现。

**验证范围与局限**：验证范围包括LIBERO、两个VLA骨干、仿真基准和真机任务。

## 英文摘要 (Abstract)

Vision-language-action (VLA) models increasingly rely on diffusion- or flow-matching-based action heads to generate continuous robot actions. These action heads typically process the denoising trajectory in a largely uniform manner. However, we observe that the conditioning focus naturally shifts across denoising stages: early stages combine language instructions and visual observations to establish a coarse action trajectory, whereas later stages place greater emphasis on current visual observations for action alignment. Based on this insight, we introduce StairVLA, a stage-aware hierarchical action generation framework that uses partially denoised actions as a natural interface between coarse long-horizon action generation and local refinement. A high-level VLA performs early denoising to produce a reusable long-horizon partially denoised action trajectory, while a lightweight refiner operates at a higher frequency to refine local action chunks using the latest observations. This design amortizes expensive high-level VLA computation while preserving frequent closed-loop correction. On LIBERO, our GR00T-style instantiation improves average success from 96.5% to 97.8% while reducing amortized inference latency from 115.0 ms to 44.2 ms per action chunk. More broadly, across two VLA backbones, simulation benchmarks, and real-robot tasks, StairVLA consistently reduces inference cost while maintaining strong task performance.

作者：Shangyuan Yuan, Xinda Qi, Yujiang Pu, Wenliang Guo, Xiaobo Tan
:::

## 原始参考来源 (Sources)

- [arXiv 2610.07756：StairVLA: Stage-Aware Hierarchical Action Generation for Vision-Language-Action Models](https://arxiv.org/abs/2610.07756)