外观
论文arXiv 2610.1140120 天前
WAM-Cache：面向高效 World Action Models 的陈旧度有界 KV 复用
原文链接：WAM-Cache: Staleness-Bounded KV Reuse for Efficient World Action Models（arXiv 2610.11401 · 发布日期：2026-10-08）
WAM-Cache: Staleness-Bounded KV Reuse for Efficient World Action Models
针对WAM闭环控制中视频DiT预填充开销，WAM-Cache以陈旧度有界KV复用加速，并在仿真和真机验证。
核心要点
- WAM闭环控制每个动作块运行视频DiT，预填充KV成为主要计算开销。
- 方法跨动作块保留分层KV，仅重算稀疏刷新token，且无需训练。
- 刷新集合结合动作专家交叉注意力与视觉潜在惊讶，并加入严格年龄上界。
- 在Fast-WAM上，视频DiT预填充FLOPs降低32-42%，精度接近密集策略。
验证范围与局限：验证范围包括RoboTwin 2.0、LIBERO仿真和真实机器人实验。
英文摘要 (Abstract)
World Action Models (WAMs) enable generalist robot manipulation by conditioning an action expert on representations from a pretrained video Diffusion Transformer (DiT). In closed-loop control, the video DiT runs at every chunk to encode the current observation into layerwise key-value (KV) pairs that the action expert queries. This prefill dominates the per-chunk computational cost, yet existing training-free accelerations leave it fully dense. We present WAM-Cache, a training-free framework that retains layerwise key-value representations across chunks and recomputes only a sparse refresh set of tokens. Crucially, we find that the intuitive heuristic of refreshing visually drifted tokens plateaus far below the dense baseline, even with an oracle predicting ground-truth KV drift. Downstream action accuracy is instead governed by where the action expert attends, not by what moved. WAM-Cache therefore selects the refresh set by uniting the action expert's cross-attention with visual latent surprise, complemented by a strict age bound that suppresses compounding error. On Fast-WAM, WAM-Cache cuts video DiT prefill FLOPs by 32-42% across RoboTwin 2.0, LIBERO, and real-world experiments, while staying within 0.7-1.8 percentage points of the dense policy in simulation and 2.5 points on a real robot.
作者：Kai Ding, Yang He, Ruijie Quan, Yi Yang
来源
本页事实对应的公开页面，以原文为准。