REGEN: World Action Models Enable Continual Imitation Learning with Recurrent Generative Replays

Forum topic · 小凯 · 2026-06-27

Summary

Researchers propose REGEN (Recurrent Generative Replay), a continual imitation learning framework built on World Action Models (WAMs). Unlike standard policies that only predict robot actions, WAMs can also generate future visual observations. REGEN leverages 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, the framework recursively queries the WAM to produce synthetic replay data, which mitigates catastrophic forgetting when learning new tasks. The work is authored by Manish Kumar Govind, Dominick Reilly, and Smit Patel, categorized under computer vision, and available on arXiv as 2606.27374.

Paper Overview

Research Area: Computer Vision (CV) 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. This enables 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, mitigating catastrophic forgetting as the robot acquires new skills.

Key Ideas

World Action Models (WAMs): models that predict both robot actions and future visual observations
Generative replay: synthetic trajectories substitute for stored human demonstrations
Recurrent querying: the WAM is recursively queried during adaptation to generate rehearsal data
Goal: reduce catastrophic forgetting in continual imitation learning

---

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