Research map / Multi-agent systems and games / Cooperative multi-agent RL
MangoBench: A Benchmark for Multi-Agent Goal-Conditioned Offline Reinforcement Learning
Yi Wang, Ningze Zhong, Zhiheng Fu, Longguang Wang, Ye Zhang, Yulan Guo
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 fully decentralized IHIQL instead of CTDE HIQL-CTDE (ablation) | hierarchical goal-conditioned MARL · AntMaze-navigate large (2x4) | success rate (%): 44.0 -> 85.4 | Improved |
| use fully decentralized IHIQL instead of CTDE HIQL-CTDE (ablation) | hierarchical goal-conditioned MARL · AntMaze-navigate giant (2x4) | success rate (%): 1.4 -> 57.3 | Improved |
| use CTDE instead of decentralized training for flat GCIVL variant (ablation) | IGCIVL · AntMaze-navigate large (2x4) | success rate (%): 15.0 -> 20.5 | No clear effect |
| use CTDE instead of decentralized training for flat GCIVL variant (ablation) | IGCIVL · AntMaze-navigate medium (2x4) | success rate (%): 76.0 -> 75.0 | No clear effect |
| evaluate with multiple goals instead of a single goal | IHIQL · lift-barrier | success rate: 78% -> 82% | Improved |
| evaluate with multiple goals instead of a single goal | GCMBC · lift-barrier | success rate: 22% -> 47% | Improved |
| evaluate with multiple goals instead of a single goal | ICRL · lift-barrier | success rate: 37% -> 56% | Improved |
| use shaped dense reward instead of sparse reward (ablation) | OMIGA · antmaze-medium-navigate | success rate | Improved |
| add goal-conditioning to offline MARL with sparse reward (GCOMIGA, GCOMAR) | OMIGA and OMAR · MangoBench locomotion tasks | success rate | Hurt |
| use vision observations instead of state observations (ablation) | goal-conditioned MARL baselines · manipulation tasks | success rate | Improved |
| use multi-agent IHIQL instead of single-agent HIQL | IHIQL · teleport mazes | success rate | Improved |
| use multi-agent baselines instead of single-agent for Ant-Soccer | multi-agent baselines · Ant-Soccer | success rate | Hurt |
Closest papers by meaning
- Dispelling the Mirage of Progress in Offline MARL through Standardised Baselines and Evaluation (NeurIPS 2024 · similarity 0.91)
- Offline Multi-Agent Reinforcement Learning via In-Sample Sequential Policy Optimization (AAAI 2025 · similarity 0.90)
- ComaDICE: Offline Cooperative Multi-Agent Reinforcement Learning with Stationary Distribution Shift Regularization (ICLR 2025 · similarity 0.90)
- OGBench: Benchmarking Offline Goal-Conditioned RL (ICLR 2025 · similarity 0.89)
- Exploiting Structure in Offline Multi-Agent RL: The Benefits of Low Interaction Rank (ICLR 2025 · similarity 0.88)
- Offline Actor-Critic Reinforcement Learning Scales to Large Models (ICML 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.
