StarCraft II micromanagement benchmark (SMAC)

The StarCraft II micromanagement benchmark (SMAC) is a dataset used to evaluate cooperative multi-agent reinforcement learning by providing a set of highly cooperative tasks that require agents to balance self-exploration and team collaboration.

Papers using StarCraft II micromanagement benchmark (SMAC) (10)

Tackling Uncertainties In Multi-agent Reinforcement Learning Through Integration Of Agent Termination Dynamics2025 · 1 citesTackling Uncertainties in Multi-Agent Reinforcement Learning through Integration of Agent Termination Dynamics2025 · 1 citesContext-aware Sparse Deep Coordination Graphs2021 · 6 citesRODE: Learning Roles to Decompose Multi-Agent Tasks2020 · 62 citesOff-Policy Multi-Agent Decomposed Policy Gradients2020 · 42 citesROMA: Multi-Agent Reinforcement Learning with Emergent Roles2020 · 32 citesRegularized Softmax Deep Multi-Agent $Q$-Learning2021 · 16 citesUneVEn: Universal Value Exploration for Multi-Agent Reinforcement Learning2020 · 7 citesSelf-Motivated Multi-Agent Exploration2023 · 2 citesContainerized Distributed Value-Based Multi-Agent Reinforcement Learning2021