# 面向基于流的视觉-语言-动作模型的多教师在策略蒸馏

> 原文链接：https://robotics.hellobike.com/papers/2026-10-15-2610-21522
> 频道：papers | 更新日期：2026-10-15 | 标签：arxiv, cs.RO
> 核心摘要：针对多领域VLA技能难以整合，提出两级在策略蒸馏，并在LIBERO-40、CALVIN和Meta-World MT50验证。

# 面向基于流的视觉-语言-动作模型的多教师在策略蒸馏

::: v-pre
> 原文链接：[Multi-Teacher On-Policy Distillation for Flow-Based Vision-Language-Action Models](https://arxiv.org/abs/2610.21522)（arXiv 2610.21522 · 发布日期：2026-10-15）

**Multi-Teacher On-Policy Distillation for Flow-Based Vision-Language-Action Models**

针对多领域VLA技能难以整合，提出两级在策略蒸馏，并在LIBERO-40、CALVIN和Meta-World MT50验证。

## 核心要点

- 专用VLA策略具备领域技能，但整合为共享策略并部署时存在分布不匹配。
- 方法结合学生驱动的环境交互与学生动作流轨迹上的教师速度监督。
- 多个冻结领域专家共同监督一个学生，更新后重新收集经验和动作流轨迹。
- 相较联合SFT初始化及其他蒸馏基线，该方法在三个基准上取得更高成功率。

**验证范围与局限**：验证范围为LIBERO-40、CALVIN五指令序列和Meta-World MT50；摘要未说明真机实验。

## 英文摘要 (Abstract)

Specialized vision-language-action (VLA) policies acquire strong domain-specific skills, yet consolidating these skills into one shared policy remains challenging. On-policy distillation (OPD) transfers teacher knowledge on inputs induced by the student, addressing the distribution mismatch encountered during deployment. For flow-based VLAs, this mismatch arises at two levels: environment observations and intermediate action-generation states. We propose a two-level on-policy distillation framework that couples student-driven environment interaction with teacher velocity supervision along student-generated action flow trajectories. Frozen domain specialists jointly supervise one student. After each update, the student collects new environment experience and generates fresh action-flow trajectories. Starting from a joint supervised fine-tuning (SFT) initialization, Velocity Loss OPD yields absolute success-rate gains of 22.55% on LIBERO-40, 33.4% on CALVIN five-instruction sequences, and 22.6% on Meta-World MT50. Compared with Action Loss OPD and Teacher-rollout SFT, it achieves absolute success-rate gains of 6.4-9.5% on LIBERO-40, 9.7-18.3% on CALVIN, and 3.2-4.2% on Meta-World MT50. Analysis by sequence length further shows larger success-rate gains on longer CALVIN sequences. These results demonstrate effective multi-domain capability consolidation through two-level on-policy supervision, with a single student deployed across domains.

作者：Lichao Ni, Yuhan Liu, Pengfei Li, Guo Hua Fang
:::

## 原始参考来源 (Sources)

- [arXiv 2610.21522：Multi-Teacher On-Policy Distillation for Flow-Based Vision-Language-Action Models](https://arxiv.org/abs/2610.21522)