# VideoVLA：视频生成模型转型机器人操作VLA实现跨具身泛化

> 原文链接：https://robotics.hellobike.com/papers/2025-12-07-2512-06963-c45a0581
> 频道：papers | 更新日期：2025-12-07 | 标签：auto, arxiv, cs.RO
> 核心摘要：研究提出VideoVLA框架，将视频生成模型转化为机器人操作VLA。基于多模态Diffusion Transformer联合预测动作序列与未来视觉结果，验证了高质量视觉想象与任务成功率正相关，并实现跨具身技能模仿与新物体操作泛化。

# VideoVLA：视频生成模型转型机器人操作VLA实现跨具身泛化

::: v-pre
> 本文由 AI 根据公开内容整理，可能有误，请以原文为准：[VideoVLA: Video Generators Can Be Generalizable Robot Manipulators](https://arxiv.org/abs/2512.06963)（arXiv 2512.06963）

## 核心内容

针对现有VLA模型在开放世界泛化能力不足的问题，arXiv论文提出VideoVLA方法，探索将大规模视频生成模型转化为通用机器人操作器。该工作核心在于利用预训练视频生成模型的先验知识，通过语言指令与当前图像输入，同时输出动作序列及对应的未来视觉预测，旨在解决新任务、新物体及新场景下的操控难题。

技术架构采用多模态Diffusion Transformer，对视频、语言和动作三种模态进行联合建模。系统利用预训练视频生成模型作为基础，执行视觉与动作的协同预测。实验表明，这种“动作+视觉后果”的双重预测策略显著提升了泛化性能，不仅能在未见过的物体上可靠执行任务，还能直接模仿其他机器人本体的操作技能，证实了视觉想象力在提升操作可靠性中的关键作用。

## 关键看点

- **双重预测机制**：同步输出动作序列与未来视觉结果，利用视觉想象质量表征动作预测可靠性
- **多模态DiT架构**：基于Diffusion Transformer联合建模视频、语言与动作，复用预训练视频生成先验
- **跨具身泛化能力**：支持模仿异构机器人本体技能，并在未见过的新物体场景中保持操作成功率
- **范式验证**：证实从视频生成到机器人操作的迁移路径有效，解锁了基于生成模型的操控泛化潜力

## 工程与产业观察

VideoVLA验证了视频生成大模型作为机器人操作基座的可行性，其“视觉想象驱动动作”的范式为具身智能提供了新思路。对于算法工程师而言，利用海量视频数据预训练带来的世界模型先验，是突破真机数据瓶颈、实现Zero-shot/Few-shot泛化的关键技术路径，值得在VLA研发中重点关注。

## 英文摘要 (Abstract)

Generalization in robot manipulation is essential for deploying robots in open-world environments and advancing toward artificial general intelligence. While recent Vision-Language-Action (VLA) models leverage large pre-trained understanding models for perception and instruction following, their ability to generalize to novel tasks, objects, and settings remains limited. In this work, we present VideoVLA, a simple approach that explores the potential of transforming large video generation models into robotic VLA manipulators. Given a language instruction and an image, VideoVLA predicts an action sequence as well as the future visual outcomes. Built on a multi-modal Diffusion Transformer, VideoVLA jointly models video, language, and action modalities, using pre-trained video generative models for joint visual and action forecasting. Our experiments show that high-quality imagined futures correlate with reliable action predictions and task success, highlighting the importance of visual imagination in manipulation. VideoVLA demonstrates strong generalization, including imitating other embodiments' skills and handling novel objects. This dual-prediction strategy - forecasting both actions and their visual consequences - explores a paradigm shift in robot learning and unlocks generalization capabilities in manipulation systems.

作者：Yichao Shen, Fangyun Wei, Zhiying Du, Yaobo Liang, Yan Lu, Jiaolong Yang, Nanning Zheng, Baining Guo
:::

## 原始参考来源 (Sources)

- [arXiv 2512.06963：VideoVLA: Video Generators Can Be Generalizable Robot Manipulators](https://arxiv.org/abs/2512.06963)