Learning to Understand Goal Specifications by Modelling Reward

TL;DR

This paper introduces a framework where reinforcement learning agents learn to follow instructions by using reward models trained on expert data, enabling better generalization and adaptation without environment-specific reward engineering.

Contribution

The authors propose a reward model-based training framework that separates instruction understanding from execution, improving generalization and adaptability in instruction-following agents.

Findings

Agents successfully learn spatial and abstract commands in grid worlds.

Reward models improve with expert data and enable adaptation to environment changes.

Framework reduces reliance on environment-specific reward engineering.

Abstract

Recent work has shown that deep reinforcement-learning agents can learn to follow language-like instructions from infrequent environment rewards. However, this places on environment designers the onus of designing language-conditional reward functions which may not be easily or tractably implemented as the complexity of the environment and the language scales. To overcome this limitation, we present a framework within which instruction-conditional RL agents are trained using rewards obtained not from the environment, but from reward models which are jointly trained from expert examples. As reward models improve, they learn to accurately reward agents for completing tasks for environment configurations---and for instructions---not present amongst the expert data. This framework effectively separates the representation of what instructions require from how they can be executed. In a simple grid world, it enables an agent to learn a range of commands requiring interaction with blocks and understanding of spatial relations and underspecified abstract arrangements. We further show the method allows our agent to adapt to changes in the environment without requiring new expert examples.