Options
Meta-world+: an improved, standardized, RL benchmark
Citation Link: https://doi.org/10.15480/882.16723
Publikationstyp
Preprint
Conference Paper
Date Issued
2025-11-21
Sprache
English
Author(s)
McLean, Reginald
Chatzaroulas, Evangelos
He, Zhanpeng
Yu, Tianhe
Julian, Ryan
Zentner, K. R.
Farsad, Nariman
Castro, Pablo Samuel
TORE-DOI
Start Page
1
End Page
21
Citation
39th Conference on Neural Information Processing Systems, NeurIPS 2025
Contribution to Conference
Publisher DOI
ArXiv ID
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.
DDC Class
005: Computer Programming, Programs, Data and Security
006: Special computer methods
Publication version
submittedVersion
Loading...
Name
2505.11289v2.pdf
Size
4.2 MB
Format
Adobe PDF