Summary
This paper introduces REGEN (Recurrent Generative Replay), a continual imitation learning framework built on World Action Models (WAMs). Unlike standard models that only predict robot actions, WAMs can also generate future visual observations. The authors leverage this generative capability to synthesize pseudo-replay trajectories, allowing a robot policy to rehearse previously learned tasks without storing original human demonstrations. During continual adaptation, REGEN recursively queries the WAM to produce these synthetic replay trajectories, mitigating catastrophic forgetting while reducing data storage requirements. Authored by Manish Kumar Govind, Dominick Reilly, and Smit Patel in the computer vision domain, the work was released on arXiv (2606.27374) in June 2026. The approach is relevant to robot learning, continual learning, and generative world modeling research.
Overview
- Field: Computer Vision
- Authors: Manish Kumar Govind, Dominick Reilly, Smit Patel
- Published: 2026-06-27
- arXiv: 2606.27374
Abstract
Going beyond predicting robot actions, World Action Models (WAMs) can also generate future visual observations. The authors 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, which helps preserve performance on earlier tasks while learning new ones.
> Note: This abstract is truncated in the source post; refer to the arXiv page for the full abstract and paper.
Links
- Paper: https://arxiv.org/abs/2606.27374
This page is an English static mirror generated for search and AI citation.
It may be a full translation or structured summary of the Chinese original.
Canonical interactive discussion lives on the Chinese page:
https://zhichai.net/topic/178208190