Nash equilibrium computation on Zero-sum games 7x7
100Success Rate (s@0.10)LP Oracle
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| LP Oracle2026.05 | 100 | — | |
| SFTSelection=s@0.10, Number of training seeds=32026.05 | 32 | — | |
| Llama-3-8B SFTBackbone=Llama-3-8B, Selection=s@0.102026.05 | 18 | — | |
| VERGESelection=s@0.10, Number of training seeds=32026.05 | 18 | — | |
| Maximin (pure)2026.05 | 10 | — | |
| Uniform2026.05 | 4 | — | |
| BaseShots=02026.05 | 0 | — | |
| Qwen3-32BValid Output Rate=8%2026.05 | 0 | — | |
| Llama-3-8B SFTBackbone=Llama-3-8B, Selection=pass@12026.05 | — | 7 | |
| SFTSelection=pass@12026.05 | — | 12 | |
| VERGESelection=pass@12026.05 | — | 7 |