Reasoning and Generalization in RL: A Tool Use Perspective

TL;DR

This paper explores how reinforcement learning agents can learn to use tools and generalize their skills across multiple tasks, inspired by animal and human tool use behaviors, and proposes a framework for measuring different types of generalization.

Contribution

It introduces a novel framework for analyzing generalization in RL through multiple transfer test sets inspired by animal tool use behaviors.

Findings

Proposes a new transfer-based evaluation method for generalization in RL.

Provides a framework inspired by the trap-tube task for analyzing tool use.

Publicly releases code for reproducibility.

Abstract

Learning to use tools to solve a variety of tasks is an innate ability of humans and has been observed of animals in the wild. However, the underlying mechanisms that are required to learn to use tools are abstract and widely contested in the literature. In this paper, we study tool use in the context of reinforcement learning and propose a framework for analyzing generalization inspired by a classic study of tool using behavior, the trap-tube task. Recently, it has become common in reinforcement learning to measure generalization performance on a single test set of environments. We instead propose transfers that produce multiple test sets that are used to measure specified types of generalization, inspired by abilities demonstrated by animal and human tool users. The source code to reproduce our experiments is publicly available at https://github.com/fomorians/gym_tool_use.