Research map / Multi-agent systems and games / Cooperative multi-agent RL
Diffusing to Coordinate: Efficient Online Multi-Agent Diffusion Policies
Zhuoran Li, Hai Zhong, Xun Wang, Qingxin Xia, Lihua Zhang, Longbo Huang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 OMAD diffusion policy instead of Gaussian HASAC for exploration | HASAC · Ant 2x4 state coverage | state coverage: 55.0% -> 68.3% | Improved |
| use OMAD instead of HATD3 | HATD3 · Ant 2x4 state coverage | state coverage: 48.4% -> 68.3% | Improved |
| increase denoising steps from 4 to 8 (ablation) | OMAD · Ant 2x4 | episode return: <4500 -> ≈6000 | Improved |
| increase denoising steps from 8 to 16 (ablation) | OMAD · Ant 2x4 | episode return: ≈6000 -> ≈6000 | No clear effect |
| increase denoising steps from 8 to 16 (cost) (ablation) | OMAD · Ant 2x4 | training time: 22h -> 26h | Hurt |
| raise V_max from low threshold 200 to about 1000 (ablation) | OMAD · Ant 2x4 | episode return | Improved |
| increase distributional atoms beyond 101 (ablation) | OMAD · Ant 2x4 | episode return | No clear effect |
| use fixed large alpha=0.1 instead of auto-tuning (ablation) | OMAD · Ant 2x4 | episode return | Hurt |
| add auto-tuned alpha versus best fixed alpha (ablation) | OMAD · Ant 2x4 | episode return: ≈6000 matches best fixed | No clear effect |
| add centralized value guidance to naive diffusion MARL extensions | MADPMD/MASDAC · MPE and MAMuJoCo | episode return | Improved |
Closest papers by meaning
- Scalable Maximum Entropy Reinforcement Learning for Diffusion Policies via Adjoint Matching (NeurIPS 2026 · similarity 0.90)
- Diffusion-based Reinforcement Learning via Q-weighted Variational Policy Optimization (NeurIPS 2024 · similarity 0.89)
- Graph Diffusion for Robust Multi-Agent Coordination (ICML 2025 · similarity 0.89)
- Revisiting Multi-Agent World Modeling from a Diffusion-Inspired Perspective (NeurIPS 2025 · similarity 0.89)
- DIME: Diffusion-Based Maximum Entropy Reinforcement Learning (ICML 2025 · similarity 0.89)
- Diffusion Actor-Critic with Entropy Regulator (NeurIPS 2024 · 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.
