Research map / Multi-agent systems and games / Cooperative multi-agent RL
Celebrating Diversity in Shared Multi-Agent Reinforcement Learning
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 |
|---|---|---|---|
| remove all intrinsic rewards (beta=0) (ablation) | CDS (QPLEX-based) · GRF academy_counterattack_hard | performance | Hurt |
| add action-aware diversity term to intrinsic reward (ablation) | CDS · GRF academy_counterattack_hard and SMAC corridor | performance | Helped |
| add L1 regularization on non-shared Q-functions (ablation) | CDS · GRF academy_counterattack_hard | performance | Helped |
| add L1 regularization on non-shared Q-functions (ablation) | CDS · SMAC corridor | performance | Helped |
| add non-shared individual Q-functions (partially shared network) instead of fully shared with identity input (ablation) | CDS · GRF academy_counterattack_hard | performance | Helped |
| add non-shared individual Q-functions instead of fully shared network (ablation) | CDS · SMAC corridor | performance | Mixed |
| add identity-aware term (H(tau|id)) to intrinsic reward (ablation) | CDS · SMAC corridor | performance | Helped |
| CDS versus QMIX, QPLEX, MAVEN, EOI on hard SMAC maps 5m_vs_6m and 3s_vs_5z | CDS · SMAC hard maps | win rate | No clear effect |
| estimate p(id|tau) with variational inference instead of assuming uniform p(id) (ablation) | CDS · SMAC 6h_vs_8z | performance | Hurt |
| use backward identity-prediction model instead of forward observation-prediction model for observation-aware diversity (ablation) | CDS · SMAC 6h_vs_8z | performance | Hurt |
Closest papers by meaning
- HyperMARL: Adaptive Hypernetworks for Multi-Agent RL (NeurIPS 2025 · similarity 0.90)
- Controlling Behavioral Diversity in Multi-Agent Reinforcement Learning (ICML 2024 · similarity 0.90)
- Enhancing Cooperative Multi-Agent Reinforcement Learning with State Modelling and Adversarial Exploration (ICML 2025 · similarity 0.89)
- Novelty-Guided Data Reuse for Efficient and Diversified Multi-Agent Reinforcement Learning (AAAI 2025 · similarity 0.89)
- GradPS: Resolving Futile Neurons in Parameter Sharing Network for Multi-Agent Reinforcement Learning (ICML 2025 · similarity 0.88)
- Learning Joint Behaviors with Large Variations (AAAI 2025 · similarity 0.88)
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.