EDELINE: Enhancing Memory in Diffusion-based World Models via Linear-Time Sequence Modeling

Jia-Hua Lee, Bor-Jiun Lin, Wei-Fang Sun, Chun-Yi Lee

Keywords: Model-based Reinforcement Learning, Atari 100k, Doom, Crafter, MAMBA, Diffusion, World Model

TL;DR: EDELINE combines diffusion models with state space models to create a world model for reinforcement learning that overcomes memory limitations in previous approaches.

Abstract:

World models represent a promising approach for training reinforcement learning agents with significantly improved sample efficiency. While most world model methods primarily rely on sequences of discrete latent variables to model environment dynamics, this compression often neglects critical visual details essential for reinforcement learning. Recent diffusion-based world models condition generation on a fixed context length of frames to predict the next observation, using separate recurrent neural networks to model rewards and termination signals. Although this architecture effectively enhances visual fidelity, the fixed context length approach inherently limits memory capacity.
In this paper, we introduce EDELINE, a unified world model architecture that integrates state space models with diffusion models. Our approach outperforms existing baselines across visually challenging Atari 100k tasks, memory-demanding Crafter benchmark, and 3D first-person ViZDoom environments, demonstrating superior performance in all these diverse challenges. Code is available at https://github.com/LJH-coding/EDELINE.

Primary Area: Reinforcement learning (e.g., decision and control, planning, hierarchical RL, robotics)

Submission Number: 5428