Horizon Imagination: Efficient On-Policy Rollout in Diffusion World Models

TL;DR

Horizon Imagination introduces an efficient on-policy diffusion-based world model for reinforcement learning that denoises multiple future observations in parallel, reducing computational costs while maintaining high control performance.

Contribution

The paper presents Horizon Imagination, a novel on-policy diffusion process with stabilization and a new sampling schedule, enabling efficient parallel denoising in world models.

Findings

Maintains control performance with half the denoising steps.

Achieves superior generation quality under varied schedules.

Reduces computational costs in diffusion world models.

Abstract

We study diffusion-based world models for reinforcement learning, which offer high generative fidelity but face critical efficiency challenges in control. Current methods either require heavyweight models at inference or rely on highly sequential imagination, both of which impose prohibitive computational costs. We propose Horizon Imagination (HI), an on-policy imagination process for discrete stochastic policies that denoises multiple future observations in parallel. HI incorporates a stabilization mechanism and a novel sampling schedule that decouples the denoising budget from the effective horizon over which denoising is applied while also supporting sub-frame budgets. Experiments on Atari 100K and Craftium show that our approach maintains control performance with a sub-frame budget of half the denoising steps and achieves superior generation quality under varied schedules. Code is available at https://github.com/leor-c/horizon-imagination.