MMaDA-VLA：具备统一多模态指令与生成能力的大型扩散式视觉-语言-行动模型

摘要

视觉-语言-动作（VLA）模型旨在通过视觉观察和自然语言指令来控制机器人执行操作任务。然而，现有的分层与自回归范式常引入冗余架构，存在时序不一致性和长周期误差累积问题，且缺乏无需额外模块即可捕捉环境动态的机制。为此，我们提出MMaDA-VLA——一个完全原生预训练的大型扩散VLA模型，将多模态理解与生成统一于单一框架。其核心在于采用原生离散扩散建模方法，将语言、图像及连续机器人控制嵌入统一离散标记空间，并通过掩码标记去噪训练单一主干网络，以并行方式联合生成未来目标观测值与动作片段。迭代去噪机制实现了全局无序优化，在无需辅助世界模型的情况下，既能提升长周期任务的一致性，又能将动作生成锚定于预测的未来视觉结果。在仿真基准与真实任务中的实验表明，该方法达到业界最优性能：在LIBERO上取得98.0%平均成功率，在CALVIN上实现4.78平均任务长度。

English

Vision-Language-Action (VLA) models aim to control robots for manipulation from visual observations and natural-language instructions. However, existing hierarchical and autoregressive paradigms often introduce architectural overhead, suffer from temporal inconsistency and long-horizon error accumulation, and lack a mechanism to capture environment dynamics without extra modules. To this end, we present MMaDA-VLA, a fully native pre-trained large diffusion VLA model that unifies multi-modal understanding and generation in a single framework. Our key idea is a native discrete diffusion formulation that embeds language, images, and continuous robot controls into one discrete token space and trains a single backbone with masked token denoising to jointly generate a future goal observation and an action chunk in parallel. Iterative denoising enables global, order-free refinement, improving long-horizon consistency while grounding actions in predicted future visual outcomes without auxiliary world models. Experiments across simulation benchmarks and real-world tasks show state-of-the-art performance, achieving 98.0% average success on LIBERO and 4.78 average length on CALVIN.

MMaDA-VLA：具备统一多模态指令与生成能力的大型扩散式视觉-语言-行动模型

MMaDA-VLA: Large Diffusion Vision-Language-Action Model with Unified Multi-Modal Instruction and Generation

摘要

Support