Investigating Pre-Training Objectives for Generalization in Vision-Based Reinforcement Learning

TL;DR

This paper introduces the Atari Pre-training Benchmark to evaluate how different pre-training objectives affect the generalization of vision-based reinforcement learning across diverse environments, highlighting the importance of task-agnostic features.

Contribution

The paper presents a new benchmark and comprehensive analysis of pre-training objectives, revealing their impact on generalization in vision-based RL.

Findings

Task-agnostic pre-training improves cross-environment generalization.

Task-specific pre-training enhances performance in similar environments.

Pre-training on diverse environments benefits overall robustness.

Abstract

Recently, various pre-training methods have been introduced in vision-based Reinforcement Learning (RL). However, their generalization ability remains unclear due to evaluations being limited to in-distribution environments and non-unified experimental setups. To address this, we introduce the Atari Pre-training Benchmark (Atari-PB), which pre-trains a ResNet-50 model on 10 million transitions from 50 Atari games and evaluates it across diverse environment distributions. Our experiments show that pre-training objectives focused on learning task-agnostic features (e.g., identifying objects and understanding temporal dynamics) enhance generalization across different environments. In contrast, objectives focused on learning task-specific knowledge (e.g., identifying agents and fitting reward functions) improve performance in environments similar to the pre-training dataset but not in varied ones. We publicize our codes, datasets, and model checkpoints at https://github.com/dojeon-ai/Atari-PB.