# RealtimeWAM：一步异步 World Action Models

> 原文链接：https://robotics.hellobike.com/papers/2026-10-05-2610-06617
> 频道：papers | 更新日期：2026-10-05 | 标签：arxiv, cs.RO
> 核心摘要：RealtimeWAM针对WAM推理低效，结合TACD一步动作生成与CEWP异步推理，在多个机器人基准和模型变体上验证。

# RealtimeWAM：一步异步 World Action Models

::: v-pre
> 原文链接：[RealtimeWAM: One-Step Asynchronous World Action Models](https://arxiv.org/abs/2610.06617)（arXiv 2610.06617 · 发布日期：2026-10-05）

**RealtimeWAM: One-Step Asynchronous World Action Models**

RealtimeWAM针对WAM推理低效，结合TACD一步动作生成与CEWP异步推理，在多个机器人基准和模型变体上验证。

## 核心要点

- 现有高效WAM虽复用视频表征，但多步动作去噪与专家顺序等待仍限制推理效率。
- TACD用冻结教师多步 rollout 终点监督，弥合局部一致性误差与最终动作精度差距。
- CEWP通过分块共享视频KV缓存重叠两个专家，仅在动作注意力消费前同步。
- 在LIBERO、LIBERO-Plus、RoboTwin等基准上性能下降小于1%，H100上约25倍加速；代码和检查点已开放。

**验证范围与局限**：验证范围为LIBERO、LIBERO-Plus、RoboTwin等基准及Fast-WAM、Faster-WAM等模型变体。

## 英文摘要 (Abstract)

World Action Models (WAMs) incorporate visual representations from video generation backbones to guide action prediction. Recent efficient WAMs adopt Mixture-of-Transformers (MoT) architectures and compute video representations once for reuse by the action expert. However, intra-expert iteration (\ie, multi-step action denoising) and inter-expert waiting (\ie, sequential execution of the video and action experts) still limit inference efficiency. To this end, we present RealtimeWAM, an extremely efficient WAM variant with one-step action generation and asynchronous inference, addressing these two bottlenecks. To reduce intra-expert iteration, we propose Teacher-Anchored Consistency Distillation (TACD) to address a local-global error gap: low local consistency error alone does not guarantee accurate final actions. TACD supplements local consistency with explicit supervision from the frozen teacher's multi-step rollout endpoint, enabling accurate one-step action generation. Additionally, we propose Cross-Expert Wavefront Pipelining (CEWP) to eliminate unnecessary expert-level waiting. It overlaps the two experts through block-wise sharing of the video KV cache, synchronizing only immediately before the corresponding action attention consumes it. Extensive experiments across diverse benchmarks (\eg, LIBERO, LIBERO-Plus and RoboTwin) and model variants (\eg, Fast-WAM and Faster-WAM) demonstrate the superiority of RealtimeWAM. Notably, RealtimeWAM maintains near-lossless performance (\ie, $<1\%$ drop) across these benchmarks while delivering significant end-to-end speedup (\eg, $\sim25\times$ on H100). Our code and checkpoints are available via this \href{https://github.com/ModelTC/LightX2V/tree/main/scripts/realtimewam}{link}.

作者：Chengtao Lv, Jinyang Du, Shuyi Feng, Yang Yong, Shiqiao Gu, Shunzi Yang, Ruihao Gong, Shen Ren, Tianwei Zhang, Wenya Wang
:::

## 原始参考来源 (Sources)

- [arXiv 2610.06617：RealtimeWAM: One-Step Asynchronous World Action Models](https://arxiv.org/abs/2610.06617)