论文HF Daily Papers · arXiv 2610.358508 天前

OSWAM单步扩散实现具身控制40ms实时推理

OSWAM通过想象Token初始化与特权信息蒸馏，将世界动作模型推理压缩至单步扩散。在RTX 5090上延迟低至40ms，提速3倍且性能持平多步基线，解决具身实时部署瓶颈。

AI 整理

具身模型
端侧部署

本文由 AI 根据公开内容整理，可能有误，请以原文为准：OSWAM: An Effective World Action Model with One-Step Action Generation（HF Daily Papers · arXiv 2610.35850）

核心内容

针对直接策略世界动作模型（WAM）因需数十步去噪导致推理延迟高、难以实时部署的问题，研究提出OSWAM框架。该方法通过引入结构化动作先验替代随机噪声初始化，并结合未来条件特权信息蒸馏，成功将高质量动作生成压缩至单次扩散步骤。实验表明，OSWAM在自动驾驶与具身控制任务中性能持平甚至优于传统多步WAM及Joint-WAM基线。

技术核心在于构建“想象Token”为扩散过程提供物理先验与动作条件表征，避免多步轨迹坍缩；同时利用特权信息蒸馏增强未来感知能力。在NVIDIA RTX 5090 GPU上，OSWAM支持实时编译推理，延迟低至40ms，相比直接策略WAM基线实现约3倍加速，验证了单步生成在高性能具身控制中的工程可行性，项目代码已开源。

关键看点

单步生成机制：利用想象Token提供物理先验初始化扩散，替代随机噪声，避免减少步数导致的动作质量坍缩
实时推理性能：在NVIDIA RTX 5090上实现40ms超低延迟，支持编译推理，较直接策略WAM基线提速约3倍
特权信息蒸馏：引入未来条件蒸馏模块增强动作生成的未来感知能力，确保单步输出具备长时序一致性
跨域验证效果：在自动驾驶与具身控制双场景中，单步OSWAM性能持平或优于需数十步的Direct-policy WAM

工程与产业观察

OSWAM证明了通过结构化先验初始化可将扩散策略压缩至单步而不损性能，40ms延迟突破了WAM实时部署瓶颈。该范式为具身智能端侧高频控制提供了新路径，显著降低了对算力与多步采样的依赖，具有极高的工程落地参考价值。

英文摘要 (Abstract)

Direct-policy World Action Models (WAM) have demonstrated the benefits of decoupling action generation from explicit future visual prediction for efficient inference in autonomous driving and embodied control. However, they still require tens of denoising steps to achieve competitive action quality, incurring substantial latency and hindering real-time deployment. Simply reducing the number of denoising steps will collapse the nonlinear multi-step denoising trajectory into coarse updates, substantially degrading action quality. To reduce the denoising steps while enabling real-time action generation, we present an effective world action model with one-step action generation, namely OSWAM, for autonomous driving and embodied control. We argue that it is the absence of structured action priors in random noise initialization that forces direct-policy WAMs to recover coherent action sequences through iterative denoising. Hence, we construct imagination tokens to initialize the diffusion process, providing physical priors and action-conditioned representations. We further introduce privileged information distillation with future conditioning to enhance future-aware action generation. As a result, OSWAM can efficiently and effectively generate high quality actions in just one diffusion step. Our experiments demonstrate that OSWAM achieves comparable or even better results in both autonomous driving and embodied control than direct-policy WAM and Joint-WAM that require tens of steps. It obtains approximately 3times speedup over direct-policy WAM baselines and supports real-time compiled inference with latency as low as 40ms on an NVIDIA RTX 5090 GPU. The project page is available at https://gwenzhang.github.io/OSWAM/

Hugging Face Daily Papers 热度：6 upvotes 作者：Guowen Zhang, Liyi Chen, Xu Hu, Pengfei Wang, Jingfan Chen, Chenhang He, Lei Zhang