# 用于 World Action Model 的无数据持续学习与 Rectified Generative Replay

> 原文链接：https://robotics.hellobike.com/papers/2026-10-13-2610-18038
> 频道：papers | 更新日期：2026-10-13 | 标签：arxiv, cs.RO
> 核心摘要：针对 WAM 顺序微调遗忘且旧数据不可用，提出 R^2-WAM 自生成并校正回放，在 LIBERO 与真实任务上验证。

# 用于 World Action Model 的无数据持续学习与 Rectified Generative Replay

::: v-pre
> 原文链接：[Data-Free Continual Learning for World Action Model with Rectified Generative Replay](https://arxiv.org/abs/2610.18038)（arXiv 2610.18038 · 发布日期：2026-10-13）

**Data-Free Continual Learning for World Action Model with Rectified Generative Replay**

针对 WAM 顺序微调遗忘且旧数据不可用，提出 R^2-WAM 自生成并校正回放，在 LIBERO 与真实任务上验证。

## 核心要点

- WAM 可快速适应新机器人任务，但顺序微调会遗忘先前任务。
- 方法让 WAM 自回归再生成旧任务轨迹，并用评估器截断完成后的无目的运动。
- 同时过滤动作与观测物理不一致的轨迹，缓解开环生成中的动作漂移。
- 在四个 LIBERO 套件和五个真实任务上，相比顺序微调遗忘降低超过90%，并接近 ER。

**验证范围与局限**：验证范围为四个 LIBERO 套件和五个真实世界任务。

## 英文摘要 (Abstract)

World action models (WAMs) built on pretrained video generation models adapt quickly to new robotic tasks, but they forget previously learned tasks under sequential fine-tuning. Experience replay (ER) mitigates such forgetting by rehearsing data of previous tasks, yet this data is often unavailable in robotics, where a policy is deployed as weights only and its training data is not shared along with it. In this work, we ask whether a WAM can achieve data-free continual learning by regenerating its own past. Our key observation is that a WAM can indeed autoregressively regenerate complete rollouts of previously learned tasks and replay them to mitigate forgetting. However, replaying such raw rollouts still lags far behind replaying real data, which we attribute to two issues. First, the WAM is unable to recognize task completion and keeps generating purposeless motion after the task is done. Second, under open-loop generation, the generated actions gradually drift away from the generated observations. To address both issues, we present rectified generative replay (R$^2$-WAM), which truncates generated rollouts at task completion with an agentic evaluator and filters out rollouts whose actions are physically inconsistent with the observations. On the four LIBERO suites and five real-world tasks, R$^2$-WAM reduces forgetting by over 90% relative to sequential fine-tuning and nearly matches ER without accessing any data from previous tasks. Our analysis further reveals that forgetting in WAMs concentrates in inverse dynamics modeling: a forgotten WAM still produces visually plausible observations of previous tasks while its generated actions fail. Project page: https://kaipengfang.github.io/r2-wam/

作者：Kaipeng Fang, Hangxu Liu, Chaojian Shi, Yunsong Zhou, Hongzi Zhu, Minyi Guo, Jingkuan Song, Lianli Gao
:::

## 原始参考来源 (Sources)

- [arXiv 2610.18038：Data-Free Continual Learning for World Action Model with Rectified Generative Replay](https://arxiv.org/abs/2610.18038)