Research map / Multi-agent systems and games / Games and equilibria
PerfectDou: Dominating DouDizhu with Perfect Information Distillation
Guan Yang, Minghuan Liu, Weijun Hong, Weinan Zhang, Fei Fang, Guangjun Zeng, Yue Lin
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 |
|---|---|---|---|
| train value function with perfect-information features (distillation) instead of imperfect-only (ablation) | PerfectDou (1e9) · DouDizhu vs DouZero 1e10 | WP / ADP: WP 0.486 -> 0.524; ADP -0.057 -> 0.014 | Improved |
| add node-level perfect reward (ablation) | PerfectDou (1e9) · DouDizhu vs DouZero 1e9 | ADP: 0.490 -> 1.270 | Improved |
| add node-level perfect reward (ablation) | PerfectDou (1e9) · DouDizhu vs DouZero 1e9 | WP: 0.738 -> 0.732 | No clear effect |
| add perfect value features and node reward to vanilla PPO (ablation) | PPO actor-critic · DouDizhu vs DouZero 1e9 | WP: 0.509 -> 0.732 | Improved |
| use PerfectDou's imperfect features in DouZero (ablation) | DouZero · DouDizhu vs PerfectDou | performance | No clear effect |
| PerfectDou 1e9 steps vs DouZero ~1e10 steps | PerfectDou · DouDizhu 100k decks | WP: 0.524 vs 0.5 baseline; ADP 0.014 | Improved |
| train longer: 1e9 to 2.5e9 steps | PerfectDou · DouDizhu vs DouZero ~1e10 | ADP: 0.014 -> 0.130 | Improved |
Later papers that compared with this method
1 later accepted paper reports PerfectDou as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- GAE Falls Short in Imperfect-Information Self-Play Reinforcement Learning (NeurIPS 2026 · 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
- DouZero: Mastering DouDizhu with Self-Play Deep Reinforcement Learning (ICML 2021 · similarity 0.87)
- Mastering Zero-Shot Interactions in Cooperative and Competitive Simultaneous Games (ICML 2024 · similarity 0.85)
- A Policy-Gradient Approach to Solving Imperfect-Information Games with Best-Iterate Convergence (ICLR 2025 · similarity 0.84)
- Learning Imperfect Information Extensive-form Games with Last-iterate Convergence under Bandit Feedback (ICML 2025 · similarity 0.84)
- GAE Falls Short in Imperfect-Information Self-Play Reinforcement Learning (NeurIPS 2026 · similarity 0.84)
- The Update-Equivalence Framework for Decision-Time Planning (ICLR 2024 · similarity 0.84)
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.
