Research map / Multi-agent systems and games / Cooperative multi-agent RL
Offline Multi-Agent Reinforcement Learning with Knowledge Distillation
Wei-Cheng Tseng, Tsun-Hsuan Wang, Yen-Chen Lin, Phillip Isola
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 |
|---|---|---|---|
| replace conventional KL distillation with relational distillation (ablation) | Ours · Equal Space | average return: −5.43 ± 0.51 -> −2.43 ± 0.72 | Improved |
| replace conventional KL distillation with relational distillation (ablation) | Ours · Highway | average return: 18.33 ± 1.27 -> 23.35 ± 0.91 | Improved |
| share weights between teacher and student mapping networks (M = N) (ablation) | Ours · Highway | average return: 23.35 ± 0.91 -> 14.92 ± 1.87 | Hurt |
| remove the mapping network (ablation) | Ours · Highway | average return: 23.35 ± 0.91 -> 18.21 ± 1.99 | Hurt |
| remove momentum update of teacher mapping network (ablation) | Ours · Highway | average return: 23.35 ± 0.91 -> 18.62 ± 1.98 | Hurt |
| remove KL divergence term (ablation) | Ours · Highway | average return: 23.35 ± 0.91 -> 20.11 ± 1.88 | Hurt |
| remove KL divergence term (ablation) | Ours · Grid-World | average return: 2.09 ± 0.22 -> 1.99 ± 0.51 | No clear effect |
| add structural relational distillation over independent decision transformers | Ours vs IDT · Highway | average return: 18.71 ± 1.53 -> 23.35 ± 0.91 | Improved |
| degrade offline dataset quality from good to poor | Ours · Highway | average return: 23.35 ± 0.91 -> 17.13 ± 1.12 | Hurt |
| degrade offline dataset quality from good to poor | Ours · Grid-World | average return: 2.09 ± 0.22 -> 1.81 ± 0.19 | Hurt |
| use sequence modeling (MADT) instead of offline RL (MA-ICQ) | baselines · Highway | average return: 18.01 ± 1.27 -> 18.78 ± 1.27 | No clear effect |
| online MAPPO finetuning of the pre-trained policy | Ours · finetuning curves | return | Improved |
Closest papers by meaning
- Offline Multi-Agent Reinforcement Learning via In-Sample Sequential Policy Optimization (AAAI 2025 · similarity 0.92)
- Learning from Good Trajectories in Offline Multi-Agent Reinforcement Learning (AAAI 2023 · similarity 0.91)
- Dispelling the Mirage of Progress in Offline MARL through Standardised Baselines and Evaluation (NeurIPS 2024 · similarity 0.91)
- ComaDICE: Offline Cooperative Multi-Agent Reinforcement Learning with Stationary Distribution Shift Regularization (ICLR 2025 · similarity 0.91)
- Cooperative Policy Agreement: Learning Diverse Policy for Offline MARL (AAAI 2025 · similarity 0.90)
- Exploiting Structure in Offline Multi-Agent RL: The Benefits of Low Interaction Rank (ICLR 2025 · similarity 0.89)
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.
