Cross-environment Cooperation Enables Zero-shot Multi-agent Coordination

Kunal Jha, Wilka Carvalho, Yancheng Liang, Simon Du, Max Kleiman-Weiner, Natasha Jaques

International Conference on Machine Learning 2025 · Oral

This talk, presented by Kunal Jha at ICML 2025, introduces **Cross-Environment Cooperation (CEC)**, a novel paradigm designed to tackle the formidable challenge of **zero-shot multi-agent coordination**. The core problem addressed is the inability of current multi-agent reinforcement learning (MARL) systems to robustly collaborate with unfamiliar partners on tasks they have never encountered, especially when the operational environment itself is new. While AI has made significant strides in single-agent and zero-sum multi-agent scenarios, achieving general collaboration—where AI can work flexibly with many people across diverse, unseen problems—remains a critical bottleneck for deploying truly intelligent agents in real-world applications like robotics, healthcare, or personal assistance.

AI review

CEC is a competent empirical contribution to zero-shot multi-agent coordination that proposes a sensible and practically motivated training regime — self-play across procedurally diverse environments — and demonstrates meaningful improvements in both cross-play metrics and human-AI interaction studies. The work is honest about its limitations, the human study is unusually careful for this subfield, and the empirical game theory framing is a reasonable tool for ruling out the 'collapsed convention' confound. However, the paper is fundamentally an empirical paper making an empirical claim, and…