Research map / Multi-agent systems and games / Games and equilibria
A Unified Game-Theoretic Approach to Multiagent Reinforcement Learning
Marc Lanctot, Vinicius Zambaldi, Audrunas Gruslys, Angeliki Lazaridou, Karl Tuyls, Julien Perolat, David Silver, Thore Graepel
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 DCH rather than InRL | Multiagent policies · Laser Tag small2 | average proportional loss (R_-): 0.342 -> 0.055 | Improved |
| Use DCH rather than InRL | Multiagent policies · Laser Tag small3 | average proportional loss (R_-): 0.625 -> 0.082 | Improved |
| Use DCH rather than InRL | Multiagent policies · Laser Tag small4 | average proportional loss (R_-): 0.717 -> 0.150 | Improved |
| Use DCH rather than InRL | Multiagent policies · Gathering field | average proportional loss (R_-): 0.003 -> 0.007 | Hurt |
| Use DCH rather than InRL | Multiagent policies · Pathfind merge | average proportional loss (R_-): 0.022 -> < 0 | Improved |
| Use the full mixed meta-strategy rather than only the highest-level policy (ablation) | DCH · Laser Tag small2, small3, and small4 | average proportional loss (R_-): 0.147, 0.27, 0.118 for highest-level policies; full-strategy values 0.055, 0.082, 0.150 | Improved |
| Use level 5 rather than level 10 (ablation) | DCH · Laser Tag small4 | average proportional loss (R_-): 0.156 vs. 0.150 | No clear effect |
| Use level 3 rather than level 10 (ablation) | DCH · Laser Tag small4 | average proportional loss (R_-): 0.246 vs. 0.150 | Hurt |
| Use decoupled PRD rather than decoupled regret-matching or Exp3 (ablation) | DCH · Two-player Leduc Poker | NashConv MAUC | Improved |
| Increase the DCH level (ablation) | DCH · Two-player Leduc Poker | NashConv | Improved |
| Use PSRO/DCH rather than NFSP | Leduc Poker · Leduc Poker | NashConv | Mixed |
| Use PSRO/DCH rather than NFSP | Leduc Poker · Leduc Poker against fixed players | performance against fixed players | Improved |
Closest papers by meaning
- Multi-Agent Meta-Reinforcement Learning: Sharper Convergence Rates with Task Similarity (NeurIPS 2023 · similarity 0.90)
- Independent Policy Gradient Methods for Competitive Reinforcement Learning (NeurIPS 2020 · similarity 0.90)
- A Policy Gradient Algorithm for Learning to Learn in Multiagent Reinforcement Learning (ICML 2021 · similarity 0.90)
- Foundations of Multi-Agent Learning in Dynamic Environments: Where Reinforcement Learning Meets Strategic Decision-Making (AAAI 2025 · similarity 0.89)
- A Generalist Hanabi Agent (ICLR 2025 · similarity 0.89)
- Learning to Deliberate: Meta-policy Collaboration for Agentic LLMs with Multi-agent Reinforcement Learning (AAAI 2026 · 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.
