Flash-WAM: 世界动作模型的模态感知蒸馏

摘要

世界-动作模型（WAM）通过迭代扩散联合生成未来视频与机器人动作，在操作基准测试中表现优异，但需要数十步去噪过程，这一成本使其无法用于实时控制。步蒸馏自然成为解决方案，但现成方法在视频-动作联合设定中失效——因为视频和动作流使用不同的信噪比偏移噪声调度，训练时噪声边际分布显著不同，这种不对称性使得单模态蒸馏方法无法适应。我们提出Flash-WAM，这是一种受一致性蒸馏启发的模态感知步蒸馏框架，它为每个模态选择与其噪声环境匹配的一致性函数：基于一致性函数族的结构分析，该分析刻画了一致性边界条件下可实现的梯度缩放，为动作流的低噪声环境采用线性梯度缩放参数化，为视频流的高噪声环境采用方差保持参数化。在LingBot-VA上实例化后，Flash-WAM将推理压缩至每个模态单步。在RoboTwin 2.0上，这使每个块的延迟从8.1秒降至NVIDIA L40S上的348毫秒，实现了23倍加速，达到实时推理能力。Flash-WAM在仿真基准测试中保持了任务成功率（RoboTwin 2.0上85.5%，LIBERO上95.7%），并在宇树G1人形机器人上大幅恢复了真实世界性能（平均60%），而采用相同步数预算的朴素一致性蒸馏仅达到24%。

English

World-action models (WAMs) jointly generate future video and robot actions through iterative diffusion, achieving strong performance on manipulation benchmarks but requiring tens of denoising steps, a cost that precludes real-time control. Step distillation has emerged as the natural remedy, but off-the-shelf methods break down in the joint video-action setting because video and action streams use different SNR-shifted noise schedules and reach training with substantially different marginal noise distributions, an asymmetry that single-modality distillation methods cannot accommodate. We introduce Flash-WAM, a modality-aware step-distillation framework inspired by consistency distillation that selects the consistency function for each modality to match its noise regime: a linear-gradient-scaling parametrization for the action stream's low-noise regime, paired with a variance-preserving parametrization for the video stream's high-noise regime, grounded in a structural analysis of the consistency-function family that characterizes the achievable gradient scaling under the consistency boundary condition. Instantiated on LingBot-VA, Flash-WAM compresses inference to a single step in each modality. On RoboTwin 2.0, this reduces per-chunk latency from 8.1 seconds to 348 ms on NVIDIA L40S, a 23{times} speedup that enables real-time inference. Flash-WAM preserves task success on simulation benchmarks (85.5% RoboTwin 2.0, 95.7% LIBERO) and substantially recovers real-world performance (60% average on a Unitree G1 humanoid robot), while naive consistency distillation drops to 24% at the same step budget.