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.

翻译：视觉-语言-动作（VLA）模型旨在通过视觉观测和自然语言指令控制机器人执行操作任务。然而，现有分层式与自回归范式往往引入架构冗余，存在时间不一致性与长程误差累积问题，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——一种完全原生预训练的大型扩散VLA模型，该模型在单一框架内统一了多模态理解与生成。其核心思想在于采用原生离散扩散公式，将语言、图像和连续机器人控制信号嵌入同一离散令牌空间，并通过掩码令牌去噪训练单一骨干网络，并行生成未来目标观测与动作片段。迭代式去噪实现了全局无顺序精细优化，在提升长程一致性的同时，无需辅助世界模型即可将动作锚定于预测的未来视觉结果中。在模拟基准与真实世界任务上的实验表明，该方法达到最优性能：在LIBERO上平均成功率达98.0%，在CALVIN上平均任务长度为4.78。