Hanabi

The 'Hanabi' dataset/benchmark is a cooperative multi-agent game used to evaluate the reasoning capabilities of agents in decentralized partially observable Markov decision processes (Dec-POMDPs).

Papers using Hanabi (4)

Probing Dec-POMDP Reasoning in Cooperative MARL2026Efficient Reinforcement Learning for Zero-Shot Coordination in Evolving Games2025Reinforcement Learning for Hanabi2025A Generalist Hanabi Agent2025