The Overcooked Generalisation Challenge

Constantin Ruhdorfer, Matteo Bortoletto, Anna Penzkofer, Andreas Bulling

Keywords: Human-AI Cooperation, Unsupervised Environment Design, Multi-Agent Reinforcement Learning

TL;DR: We introduce the Overcooked Generalisation Challenge (OGC) – the first bench-mark to study reinforcement learning agents’ zero-shot cooperation abilities when faced with novel partners and levels in the Overcooked-AI environment.

Abstract:

We introduce the Overcooked Generalisation Challenge (OGC) – the first bench-mark to study reinforcement learning agents’ zero-shot cooperation abilities when faced with novel partners and levels in the Overcooked-AI environment.
This perspective starkly contrasts a large body of previous work that has evaluated cooperating agents only on the same level or with the same partner, thus failing to capture generalisation abilities essential for real-world human-AI cooperation.
Our challenge interfaces with state-of-the-art dual curriculum design (DCD) methods to generate auto-curricula for training general agents in Overcooked.
It is the first cooperative multi-agent environment specially designed for DCD methods and, consequently, the first evaluated with state-of-the-art methods.
It is fully GPU-accelerated, built on the DCD benchmark suite minimax, and freely available under an open-source license: http://anonymised.edu.
We show that state-of-the-art DCD algorithms fail to produce useful policies on this novel challenge, even if combined with recent network architectures specifically designed for scalability and generalisability.
As such, the OGC pushes the boundaries of real-world human-AI cooperation by enabling research on the impact of generalisation on cooperating agents.

Primary Area: datasets and benchmarks

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 6854