Benchmarking the Spectrum of Agent Capabilities

TL;DR

This paper introduces Crafter, a comprehensive benchmark environment in an open world survival game that evaluates diverse agent capabilities through meaningful achievements, promoting research on generalization, exploration, and reasoning.

Contribution

The paper presents Crafter, a novel unified benchmark environment for assessing a wide range of agent abilities in a single setting, reducing the need for multiple specialized benchmarks.

Findings

Crafter effectively evaluates general abilities in a complex environment.

Reward-maximizing agents exhibit sophisticated behaviors like building and resource management.

Baseline scores demonstrate the benchmark's suitability for future research.

Abstract

Evaluating the general abilities of intelligent agents requires complex simulation environments. Existing benchmarks typically evaluate only one narrow task per environment, requiring researchers to perform expensive training runs on many different environments. We introduce Crafter, an open world survival game with visual inputs that evaluates a wide range of general abilities within a single environment. Agents either learn from the provided reward signal or through intrinsic objectives and are evaluated by semantically meaningful achievements that can be unlocked during each episode, such as discovering resources and crafting tools. Consistently unlocking all achievements requires strong generalization, deep exploration, and long-term reasoning. We experimentally verify that Crafter is of appropriate difficulty to drive future research and provide baselines scores of reward agents and unsupervised agents. Furthermore, we observe sophisticated behaviors emerging from maximizing the reward signal, such as building tunnel systems, bridges, houses, and plantations. We hope that Crafter will accelerate research progress by quickly evaluating a wide spectrum of abilities.