Synthetic Data is Sufficient for Zero-Shot Visual Generalization from Offline Data

TL;DR

This paper demonstrates that generating synthetic data via diffusion models enhances zero-shot visual generalization in offline reinforcement learning, reducing overfitting and improving performance without altering existing algorithms.

Contribution

It introduces a simple two-step method to augment offline data with synthetic samples using diffusion models, improving generalization in visual RL tasks.

Findings

Synthetic data improves zero-shot generalization.

Method reduces the generalization gap.

No changes needed in existing RL algorithms.

Abstract

Offline reinforcement learning (RL) offers a promising framework for training agents using pre-collected datasets without the need for further environment interaction. However, policies trained on offline data often struggle to generalise due to limited exposure to diverse states. The complexity of visual data introduces additional challenges such as noise, distractions, and spurious correlations, which can misguide the policy and increase the risk of overfitting if the training data is not sufficiently diverse. Indeed, this makes it challenging to leverage vision-based offline data in training robust agents that can generalize to unseen environments. To solve this problem, we propose a simple approach generating additional synthetic training data. We propose a two-step process, first augmenting the originally collected offline data to improve zero-shot generalization by introducing diversity, then using a diffusion model to generate additional data in latent space. We test our method across both continuous action spaces (Visual D4RL) and discrete action spaces (Procgen), demonstrating that it significantly improves generalization without requiring any algorithmic changes to existing model-free offline RL methods. We show that our method not only increases the diversity of the training data but also significantly reduces the generalization gap at test time while maintaining computational efficiency. We believe this approach could fuel additional progress in generating synthetic data to train more general agents in the future.