StarCraft Multi-Agent Challenge (SMACv-2)

The StarCraft Multi-Agent Challenge (SMACv-2) is a benchmark used to evaluate cooperative multi-agent reinforcement learning algorithms in complex, partially observable environments.

Papers using StarCraft Multi-Agent Challenge (SMACv-2) (9)

Smacv2: An Improved Benchmark For Cooperative Multi-agent Reinforcement Learning2022 · 26 citesSmac-hard: Enabling Mixed Opponent Strategy Script And Self-play On SMAC2024Low-rank Agent-specific Adaptation (lorasa) For Multi-agent Policy Learning2025Closed-Loop Vision-Language Planning for Multi-Agent Coordination2025Enabling Multi-Agent Transfer Reinforcement Learning via Scenario Independent Representation2024 · 4 citesSemi-On-Policy Training for Sample Efficient Multi-Agent Policy Gradients2021 · 1 citesMulti-Task Multi-Agent Shared Layers are Universal Cognition of Multi-Agent Coordination2023Decentralized Transformers with Centralized Aggregation are Sample-Efficient Multi-Agent World Models2024Offline-to-Online Multi-Agent Reinforcement Learning with Offline Value Function Memory and Sequential Exploration2024