Adversarial Game Playing on Two Dollar
40.62GPT-5.1 ScoreDEPT
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| DEPTBackbone=Qwen3-8B-Base2026.05 | 40.62 | 38.02 | 30.86 | 36.5 | |
| MARSBackbone=Qwen3-8B-Base2026.05 | 35.16 | 25.13 | 26.82 | 29.04 | |
| DEPTBackbone=Qwen3-4B-Base2026.05 | 33.2 | 36.59 | 32.42 | 34.07 | |
| GRPOBackbone=Qwen3-4B-Base2026.05 | 26.56 | 29.43 | 27.73 | 27.91 | |
| SPIRALBackbone=Qwen3-4B-Base2026.05 | 26.3 | 30.08 | 26.56 | 27.65 | |
| MARSBackbone=Qwen3-4B-Base2026.05 | 26.17 | 30.08 | 26.82 | 27.69 | |
| SPAGBackbone=Qwen3-4B-Base2026.05 | 25.91 | 24.48 | 26.17 | 25.52 | |
| SPIRALBackbone=Qwen3-8B-Base2026.05 | 25.65 | 29.17 | 23.83 | 26.22 | |
| GRPOBackbone=Qwen3-8B-Base2026.05 | 25.52 | 26.69 | 25.78 | 26 | |
| SPAGBackbone=Qwen3-8B-Base2026.05 | 17.58 | 16.93 | 23.44 | 19.31 | |
| VANILLABackbone=Qwen3-8B-Base2026.05 | 1.69 | 1.17 | 3.52 | 2.13 | |
| VANILLABackbone=Qwen3-4B-Base2026.05 | 1.43 | 0.78 | 2.47 | 1.56 |