Research map / Multi-agent systems and games / Cooperative multi-agent RL
MAVEN: Multi-Agent Variational Exploration
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 |
|---|---|---|---|
| Use MAVEN's latent-conditioned committed exploration instead of QMIX's ε-greedy exploration | QMIX · 10-step matrix game | total payoff: 10 -> 13 | Helped |
| Learn the hierarchical policy over z instead of fixing it to a uniform distribution (ablation) | MAVEN · SMAC micro_corridor | return | Helped |
| Add variational mutual-information loss to the uniform hierarchical policy (ablation) | MAVEN · SMAC micro_corridor | return | Helped |
| Use per-time-step rather than per-trajectory mutual-information loss (ablation) | MAVEN · SMAC scenarios | performance | No clear effect |
| Vary the number of discrete latent categories (ablation) | MAVEN · SMAC micro_corridor | return | Mixed |
| Use committed latent-conditioned exploration rather than ε-greedy exploration | QMIX · SMAC 2-corridors after short-corridor closure | win rate | Helped |
| Use MAVEN's latent-conditioned exploration | QMIX and other compared algorithms · SMAC zealot_cave, depths 3 and 4 | average reward | Helped |
Closest papers by meaning
- QMIX: Monotonic Value Function Factorisation for Deep Multi-Agent Reinforcement Learning (ICML 2018 · similarity 0.91)
- Weighted QMIX: Expanding Monotonic Value Function Factorisation for Deep Multi-Agent Reinforcement Learning (NeurIPS 2020 · similarity 0.89)
- UneVEn: Universal Value Exploration for Multi-Agent Reinforcement Learning (ICML 2021 · similarity 0.89)
- Q-value Path Decomposition for Deep Multiagent Reinforcement Learning (ICML 2020 · similarity 0.88)
- Learning Joint Behaviors with Large Variations (AAAI 2025 · similarity 0.87)
- Dr. MAS: Stable Reinforcement Learning for Multi-Agent LLM Systems (NeurIPS 2026 · similarity 0.87)
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.