Meta-World+: An Improved, Standardized, RL Benchmark

TL;DR

Meta-World+ is a new, standardized version of the Meta-World benchmark that ensures reproducibility, improves usability, and clarifies previous results for evaluating multi-task and meta-reinforcement learning agents.

Contribution

The paper introduces Meta-World+ with full reproducibility, enhanced ergonomics, and customizable task sets, addressing inconsistencies in previous versions.

Findings

Reproducible results across different studies

Enhanced control over task selection

Insights into benchmark design for multi-task RL

Abstract

Meta-World is widely used for evaluating multi-task and meta-reinforcement learning agents, which are challenged to master diverse skills simultaneously. Since its introduction however, there have been numerous undocumented changes which inhibit a fair comparison of algorithms. This work strives to disambiguate these results from the literature, while also leveraging the past versions of Meta-World to provide insights into multi-task and meta-reinforcement learning benchmark design. Through this process we release a new open-source version of Meta-World (https://github.com/Farama-Foundation/Metaworld/) that has full reproducibility of past results, is more technically ergonomic, and gives users more control over the tasks that are included in a task set.