Research map / Multi-agent systems and games / Cooperative multi-agent RL
Distributional Reward Estimation for Effective Multi-agent Deep Reinforcement Learning
Jifeng Hu, Yanchao Sun, Hechang Chen, Sili Huang, haiyin piao, Yi Chang, Lichao Sun
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 |
|---|---|---|---|
| add regularization term L_R (ablation) | DRE-MARL l_SMO+g_SS · CN-3 team reward | mean episodic reward: -365.0 -> -235.1 | Improved |
| add regularization term L_R (ablation) | DRE-MARL l_MO+g_SS · CN-3 team reward | mean episodic reward: -354.2 -> -252.6 | Improved |
| add regularization term L_R (ablation) | DRE-MARL l_MO+g_MO · CN-3 team reward | mean episodic reward: -258.9 -> -242.1 | Improved |
| add regularization term L_R (ablation) | DRE-MARL l_SS+g_SS · CN-3 team reward | mean episodic reward: -274.6 -> -272.7 | No clear effect |
| add distributional reward estimation with l_SMO+g_SS (ablation) | MADDPG-style baseline · CN-3 team reward | mean episodic reward: -258.2 -> -235.1 | Improved |
| use l_SS+g_SS aggregation instead of l_MO+g_MO (ablation) | DRE-MARL · CN-3 team reward | mean episodic reward: -242.1 -> -272.7 | Hurt |
| use mean-operation aggregation l_MO+g_MO instead of l_SS+g_SS (ablation) | DRE-MARL · individual rewards, r_ac-dist eval | mean episodic reward: 7.893 -> 9.360 | Improved |
| use mean-operation aggregation l_MO+g_MO instead of l_SS+g_SS (ablation) | DRE-MARL · individual rewards, no r_ac-dist eval | mean episodic reward: 7.016 -> 9.797 | Improved |
| replace DRE-MARL (l_MO+g_MO) with p2p-MARL (ablation) | DRE-MARL · individual rewards, r_ac-dist eval | mean episodic reward: 9.360 -> 7.517 | Hurt |
| multi-action-branch reward estimation vs point-to-point estimation | DRE-MARL · normalized mean over reward settings | mean episodic reward: 9.104 -> 9.802 | Improved |
| DFAC-diql(128) instead of DRE-MARL l_SS+g_SS (ablation) | DRE-MARL · CN-3 | mean episodic reward: -272.7 -> -786.9 | Hurt |
| DFAC-dmix(128) instead of DRE-MARL l_SS+g_SS (ablation) | DRE-MARL · CN-3 | mean episodic reward: -272.7 -> -1121.6 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- QMIX — QMIX: Monotonic Value Function Factorisation for Deep Multi-Agent Reinforcement Learning (ICML 2018) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- The Distributional Reward Critic Framework for Reinforcement Learning Under Perturbed Rewards (AAAI 2025 · similarity 0.88)
- O-MAPL: Offline Multi-agent Preference Learning (ICML 2025 · similarity 0.87)
- SrSv: Integrating Sequential Rollouts with Sequential Value Estimation for Multi-agent Reinforcement Learning (AAAI 2025 · similarity 0.87)
- Action Gaps and Advantages in Continuous-Time Distributional Reinforcement Learning (NeurIPS 2024 · similarity 0.87)
- Dispelling the Mirage of Progress in Offline MARL through Standardised Baselines and Evaluation (NeurIPS 2024 · similarity 0.87)
- Value-Based Deep Multi-Agent Reinforcement Learning with Dynamic Sparse Training (NeurIPS 2024 · 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.
