Adversarial Game Playing on Negotiation
17.84GPT-5.1 ScoreDEPT
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DEPTBackbone=Qwen3-4B-Base2026.05 | 17.84 | 36.46 | 43.67 | 32.35 | |
| GRPOBackbone=Qwen3-4B-Base2026.05 | 15.23 | 22.01 | 20.39 | 19.21 | |
| SPIRALBackbone=Qwen3-4B-Base2026.05 | 15.1 | 20.57 | 14.83 | 16.84 | |
| MARSBackbone=Qwen3-4B-Base2026.05 | 14.06 | 24.35 | 23.79 | 20.73 | |
| DEPTBackbone=Qwen3-8B-Base2026.05 | 14.06 | 32.81 | 48.76 | 31.88 | |
| SPAGBackbone=Qwen3-8B-Base2026.05 | 9.11 | 19.79 | 11.01 | 13.31 | |
| SPAGBackbone=Qwen3-4B-Base2026.05 | 8.98 | 22.95 | 18.35 | 16.76 | |
| SPIRALBackbone=Qwen3-8B-Base2026.05 | 8.59 | 22.27 | 21.04 | 17.3 | |
| MARSBackbone=Qwen3-8B-Base2026.05 | 8.46 | 16.8 | 23.04 | 16.1 | |
| VANILLABackbone=Qwen3-8B-Base2026.05 | 6.25 | 9.11 | 1.69 | 5.69 | |
| GRPOBackbone=Qwen3-8B-Base2026.05 | 5.47 | 17.97 | 20.68 | 14.7 | |
| VANILLABackbone=Qwen3-4B-Base2026.05 | 0.52 | 2.21 | 0.39 | 1.04 |