REGEN: World Action Models Enable Continual Imitation Learning via Recurrent Generative Replay

Forum topic · 小凯 · 2026-06-27

Summary

This paper introduces Recurrent Generative Replay (REGEN), a continual imitation learning framework built on World Action Models (WAMs). Beyond predicting robot actions, WAMs can generate future visual observations, and REGEN leverages this generative capability to synthesize pseudo-replay trajectories. During continual adaptation, the framework recursively queries the WAM to produce synthetic trajectories of previously learned tasks, allowing a robot policy to rehearse old skills without storing original human demonstrations. This mitigates catastrophic forgetting in continual learning scenarios, enabling robots to acquire new skills while retaining performance on earlier tasks. The authors—Manish Kumar Govind, Dominick Reilly, and Smit Patel—demonstrate REGEN's effectiveness across a range of robot manipulation tasks, reporting significant improvements in continual learning performance over prior approaches. The work spans computer vision and robotics and is available on arXiv (2606.27374).

Paper Overview

Research area: Computer Vision / Robotics
Authors: Manish Kumar Govind, Dominick Reilly, Smit Patel
Published: 2026-06-27
arXiv: 2606.27374

Summary

Going beyond predicting robot actions, World Action Models (WAMs) can also generate future visual observations. Building on this generative capability, the authors propose Recurrent Generative Replay (REGEN), a continual imitation learning framework that synthesizes pseudo-replay trajectories, enabling a robot policy to rehearse previously learned tasks without storing their original human demonstrations.

During continual adaptation, REGEN recursively queries the WAM to synthesize pseudo-replay trajectories, which are used to review past tasks while learning new ones. This approach mitigates catastrophic forgetting in continual learning scenarios, allowing robots to continuously acquire new skills while maintaining performance on previously learned tasks.

The authors demonstrate REGEN's effectiveness on a variety of robot manipulation tasks, achieving significant improvements in continual learning performance.

Abstract (Original)

Going beyond predicting robot actions, World Action Models (WAMs) can also generate future visual observations. We build on this generative capability to propose Recurrent Generative Replay (REGEN), a continual imitation learning framework that synthesizes pseudo-replay trajectories, enabling a robot policy to rehearse previously learned tasks without storing their original human demonstrations. During continual adaptation, REGEN recursively queries the WAM to synthesize pseudo-replay trajectories...

*Auto-collected on 2026-06-27.*

REGEN: World Action Models Enable Continual Imitation Learning via Recurrent Generative Replay

Summary

Paper Overview

Summary

Abstract (Original)

Tags