Deep reinforcement learning from human preferences

TL;DR

This paper introduces a method for training reinforcement learning agents using human preferences over trajectory pairs, enabling complex tasks to be learned efficiently without explicit reward functions.

Contribution

It demonstrates that human preferences can effectively guide RL agents in complex environments, reducing human oversight costs and enabling learning of novel behaviors.

Findings

Successfully trained agents on Atari and robot tasks using less than 1% human feedback.

Achieved complex behaviors with approximately one hour of human input.

Outperformed previous methods in learning from human preferences.

Abstract

For sophisticated reinforcement learning (RL) systems to interact usefully with real-world environments, we need to communicate complex goals to these systems. In this work, we explore goals defined in terms of (non-expert) human preferences between pairs of trajectory segments. We show that this approach can effectively solve complex RL tasks without access to the reward function, including Atari games and simulated robot locomotion, while providing feedback on less than one percent of our agent's interactions with the environment. This reduces the cost of human oversight far enough that it can be practically applied to state-of-the-art RL systems. To demonstrate the flexibility of our approach, we show that we can successfully train complex novel behaviors with about an hour of human time. These behaviors and environments are considerably more complex than any that have been previously learned from human feedback.