Reasoning Quality Evaluation on 3-player Leduc Hold'em (test)
193Hit Rate (HR)Qwen2.5-7B_ToolPoker
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen2.5-7B_ToolPokerBackbone=Qwen2.5-7B, Framework=ToolPoker2026.01 | 193 | 190 | 1.88 | 1.9 | |
| GPT-4.1-miniModel type=Baseline2026.01 | 100 | 175 | 1.83 | 1.52 | |
| Qwen2.5-7BModel type=Vanilla2026.01 | 93 | 88 | 1.6 | 1.14 |