Research map / Multi-agent systems and games / Cooperative multi-agent RL
NonZero: Interaction-Guided Exploration for Multi-Agent Monte Carlo Tree Search
Sizhe Tang, Zuyuan Zhang, Mahdi Imani, Tian Lan
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 NonZero instead of MAZero for planning | MuZero-style multi-agent MCTS · MatGame 8 agents 10 actions, non-linear, 1000 steps | mean score: 615.2 -> 637.2 | Improved |
| use NonZero instead of MAZero in the smallest non-linear MatGame | MuZero-style multi-agent MCTS · MatGame 2 agents 3 actions, non-linear, 500 steps | mean score: 49.1 -> 49.7 | No clear effect |
| use NonZero instead of MAZero in the larger linear MatGame | MuZero-style multi-agent MCTS · MatGame 4 agents 5 actions, linear, 1000 steps | mean score: 175.2 -> 186.2 | Improved |
| use NonZero instead of MAZero on SMAC maps | MuZero-style multi-agent MCTS · SMAC (3m, so_many_baneling, 2m_vs_1z) | win rate: over 96% win rate on all 3 maps; 50% to 70% fewer environment steps than MAZero | Improved |
| use NonZero instead of baselines on SMACv2 | MuZero-style multi-agent MCTS · SMACv2 (Protoss, Zerg, Terran 5_vs_5) | win rate: nearly doubles win rate vs best baseline | Improved |
| add second-order curvature estimation (vs first-order updates only) (ablation) | NonZero · MatGame (6 agents 8 actions linear; 8 agents 10 actions non-linear) | mean score | Improved |
| add hypernetwork mixing net for initializing theta (ablation) | NonZero · MatGame (6 agents 8 actions linear; 8 agents 10 actions non-linear) | mean score | Improved |
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
- MALinZero: Efficient Low-Dimensional Search for Mastering Complex Multi-Agent Planning (NeurIPS 2025 · similarity 0.92)
- Decentralized Monte Carlo Tree Search for Partially Observable Multi-Agent Pathfinding (AAAI 2024 · similarity 0.89)
- Multiagent Gumbel MuZero: Efficient Planning in Combinatorial Action Spaces (AAAI 2024 · similarity 0.88)
- WebPilot: A Versatile and Autonomous Multi-Agent System for Web Task Execution with Strategic Exploration (AAAI 2025 · similarity 0.88)
- Randomized Exploration in Cooperative Multi-Agent Reinforcement Learning (NeurIPS 2024 · similarity 0.87)
- Imagine, Initialize, and Explore: An Effective Exploration Method in Multi-Agent Reinforcement Learning (AAAI 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.
