Overcooked-AI

'Overcooked-AI' is a benchmark that contains various layouts of the cooperative cooking game "Overcooked," used to evaluate the effectiveness of reward design in multi-agent reinforcement learning scenarios characterized by different levels of complexity and interaction.

Papers using Overcooked-AI (7)

Adaptive Human-AI Coordination via Hierarchical Action Disentanglement2026Partner-Aware Hierarchical Skill Discovery for Robust Human-AI Collaboration2026Large Language Model Guided Incentive Aware Reward Design for Cooperative Multi-Agent Reinforcement Learning2026Unsupervised Partner Design Enables Robust Ad-hoc Teamwork2025Generalizable Agent Modeling for Agent Collaboration-Competition Adaptation with Multi-Retrieval and Dynamic Generation2025The Overcooked Generalisation Challenge: Evaluating Cooperation with Novel Partners in Unknown Environments Using Unsupervised Environment Design2024Towards Efficient LLM Grounding for Embodied Multi-Agent Collaboration2024 · 1 cites