What question did this study set out to answer?

This work aims to clarify discrepancies in Meta-World's benchmarks and enhance it for better evaluation of algorithms.

February 19, 20260 citationsOpen Access

Meta-world+: an improved, standardized, RL benchmark

RMReginald McLean ECEvangelos Chatzaroulas LMLuc McCutcheon

Key Points

This work aims to clarify discrepancies in Meta-World's benchmarks and enhance it for better evaluation of algorithms.
Reviewed past versions of Meta-World to assess changes and inconsistencies.
Released a new open-source version of Meta-World with improved technical design.
Ensured full reproducibility of previous results and increased user control over task sets.
Clarified existing discrepancies in the Meta-World benchmarks.
Provided a more user-friendly interface for task management.
Enhanced the reliability of multi-task and meta-reinforcement learning evaluations.

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-World1 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.

问 AI

Bookmark

View Full Paper