Research map / Multi-agent systems and games / Cooperative multi-agent RL
Counterfactual Multi-Agent Policy Gradients
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| counterfactual baseline instead of V baseline (central-QV -> COMA) (ablation) | COMA · StarCraft 3m | win rate %: 83 (5) -> 87 (3) | Helped |
| centralised critic instead of independent critic (IAC-Q -> COMA) (ablation) | COMA · StarCraft 3m | win rate %: 56 (6) -> 87 (3) | Helped |
| counterfactual baseline instead of V baseline (central-QV -> COMA) (ablation) | COMA · StarCraft 5m | win rate %: 71 (9) -> 81 (5) | Helped |
| centralised critic instead of independent critic (IAC-Q -> COMA) (ablation) | COMA · StarCraft 5m | win rate %: 58 (3) -> 81 (5) | Helped |
| counterfactual baseline instead of V baseline (central-QV -> COMA) (ablation) | COMA · StarCraft 5w | win rate %: 76 (1) -> 82 (3) | Helped |
| centralised critic instead of independent critic (IAC-Q -> COMA) (ablation) | COMA · StarCraft 5w | win rate %: 57 (5) -> 82 (3) | Helped |
| counterfactual baseline instead of V baseline (central-QV -> COMA) (ablation) | COMA · StarCraft 2d 3z | win rate %: 39 (5) -> 47 (5) | Helped |
| centralised critic instead of independent critic (IAC-Q -> COMA) (ablation) | COMA · StarCraft 2d 3z | win rate %: 19 (21) -> 47 (5) | Helped |
| learn central Q with counterfactual advantage instead of central V (ablation) | COMA · StarCraft 4 maps | win rate % | Helped |
Closest papers by meaning
- FACMAC: Factored Multi-Agent Centralised Policy Gradients (NeurIPS 2021 · similarity 0.89)
- Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environments (NeurIPS 2017 · similarity 0.89)
- Learning Implicit Credit Assignment for Cooperative Multi-Agent Reinforcement Learning (NeurIPS 2020 · similarity 0.88)
- COSAC: Counterfactual Credit Assignment in Sequential Cooperative Teams (NeurIPS 2026 · similarity 0.86)
- Agent-Centric Actor-Critic for Asynchronous Multi-Agent Reinforcement Learning (ICML 2025 · similarity 0.85)
- Finite-Time Global Optimality Convergence in Deep Neural Actor-Critic Methods for Decentralized Multi-Agent Reinforcement Learning (ICML 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.