Reward learning from human preferences and demonstrations in Atari

TL;DR

This paper introduces a method combining human demonstrations and preferences to learn reward functions for Atari games, enabling agents to outperform baselines and achieve superhuman performance without explicit game rewards.

Contribution

It presents a novel approach that integrates demonstrations and preferences for reward learning, improving performance in Atari games without relying on predefined rewards.

Findings

Outperforms imitation learning in 7 out of 9 games

Achieves superhuman performance in 2 games

Identifies issues like reward hacking and effects of label noise

Abstract

To solve complex real-world problems with reinforcement learning, we cannot rely on manually specified reward functions. Instead, we can have humans communicate an objective to the agent directly. In this work, we combine two approaches to learning from human feedback: expert demonstrations and trajectory preferences. We train a deep neural network to model the reward function and use its predicted reward to train an DQN-based deep reinforcement learning agent on 9 Atari games. Our approach beats the imitation learning baseline in 7 games and achieves strictly superhuman performance on 2 games without using game rewards. Additionally, we investigate the goodness of fit of the reward model, present some reward hacking problems, and study the effects of noise in the human labels.