Strategic Reasoning on gamma-Bench
94.87Guessing PerformanceFoPO
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| FoPOTraining Set=RSA, Backbone=Qwen3-14B2026.04 | 94.87 | 50 | 88.97 | 20.3 | 10.94 | 82.82 | 86.15 | 62.01 | |
| PPOTraining Set=RSA, Backbone=Qwen3-14B2026.04 | 94.27 | 49.67 | 81.12 | 27 | 10.21 | 78.75 | 84.12 | 60.73 | |
| ArCHerTraining Set=RSA, Backbone=Qwen3-14B2026.04 | 94.25 | 47.67 | 76.91 | 16.6 | 10.96 | 88.04 | 86.07 | 60.07 | |
| ICTTraining Set=RSA, Backbone=Qwen3-14B2026.04 | 94.24 | 36.33 | 84.24 | 13.7 | 10.61 | 74.66 | 81.85 | 56.52 | |
| FoPOTraining Set=Taboo + RSA, Backbone=Qwen3-14B2026.04 | 94.12 | 52.33 | 87.85 | 32.7 | 11.76 | 87.29 | 84.04 | 64.3 | |
| PPOTraining Set=Taboo + RSA, Backbone=Qwen3-14B2026.04 | 93.88 | 43.83 | 85.79 | 32.4 | 11.73 | 84 | 83.07 | 62.1 | |
| GRPOTraining Set=Taboo, Backbone=Qwen3-14B2026.04 | 93.85 | 34.5 | 80.33 | 22.8 | 11.31 | 89.36 | 82.41 | 59.22 | |
| FoPOTraining Set=Taboo, Backbone=Qwen3-14B2026.04 | 93.85 | 35.33 | 92.3 | 26.9 | 11.36 | 86.76 | 85.1 | 61.66 | |
| PPOTraining Set=Taboo, Backbone=Qwen3-14B2026.04 | 93.83 | 38.5 | 79.1 | 22.5 | 11.38 | 91.17 | 82.87 | 59.91 | |
| ArCHerTraining Set=Taboo + RSA, Backbone=Qwen3-14B2026.04 | 93.83 | 42 | 80.15 | 30.1 | 11.16 | 80.67 | 85.77 | 60.53 | |
| ArCHerTraining Set=Taboo, Backbone=Qwen3-14B2026.04 | 93.71 | 37 | 84.42 | 26.7 | 11.06 | 89 | 80.29 | 60.31 | |
| GR.FoPOTraining Set=Taboo, Backbone=Qwen3-14B2026.04 | 93.62 | 35.33 | 87.25 | 24.7 | 11.82 | 93.5 | 84.25 | 61.5 | |
| ICTTraining Set=Taboo + RSA, Backbone=Qwen3-14B2026.04 | 92.39 | 40.33 | 84.24 | 32.3 | 11.22 | 79.54 | 83.21 | 60.46 | |
| ICTTraining Set=Taboo, Backbone=Qwen3-14B2026.04 | 91.98 | 30.5 | 78.68 | 18.6 | 11.41 | 76.48 | 81.18 | 55.55 | |
| FoPOTraining Set=RSA, Backbone=Llama-3-8B-Instruct2026.04 | 89.59 | 67.83 | 60.36 | 89.1 | 11.3 | 29.42 | 67.4 | 59.29 | |
| ArCHerTraining Set=RSA, Backbone=Llama-3-8B-Instruct2026.04 | 87.88 | 67.17 | 58.34 | 88.1 | 10.43 | 19.35 | 66.82 | 56.87 | |
| ICTTraining Set=RSA, Backbone=Llama-3-8B-Instruct2026.04 | 86.64 | 68.17 | 51.38 | 72.2 | 10.09 | 19.08 | 67.08 | 53.52 | |
| GR.FoPOTraining Set=Taboo, Backbone=Llama-3-8B-Instruct2026.04 | 84.57 | 71.67 | 66.46 | 94.7 | 18.51 | 29.09 | 49.57 | 59.22 | |
| ArCHerTraining Set=Taboo, Backbone=Llama-3-8B-Instruct2026.04 | 84.48 | 67.33 | 65.81 | 94.7 | 16.7 | 18.05 | 48.62 | 56.53 | |
| FoPOTraining Set=Taboo, Backbone=Llama-3-8B-Instruct2026.04 | 84.3 | 69.33 | 71.65 | 95.2 | 20.84 | 24.6 | 47.96 | 59.13 | |
| ICTTraining Set=Taboo, Backbone=Llama-3-8B-Instruct2026.04 | 83.89 | 66.67 | 59.32 | 95.71 | 16.42 | 16.32 | 44.4 | 54.68 | |
| PPOTraining Set=Taboo, Backbone=Llama-3-8B-Instruct2026.04 | 83.25 | 67.17 | 69.75 | 95.3 | 17.31 | 17.67 | 46.25 | 56.67 | |
| PPOTraining Set=RSA, Backbone=Llama-3-8B-Instruct2026.04 | 83.16 | 67.67 | 55.23 | 89.7 | 10.41 | 21.57 | 65.35 | 56.16 | |
| GRPOTraining Set=Taboo, Backbone=Llama-3-8B-Instruct2026.04 | 82.99 | 69 | 68.82 | 96.1 | 20.31 | 18.78 | 47.4 | 57.63 | |
| FoPOTraining Set=Taboo + RSA, Backbone=Llama-3-8B-Instruct2026.04 | 80.47 | 72.83 | 64.61 | 98.4 | 13.27 | 32.94 | 58.05 | 60.08 | |
| ArCHerTraining Set=Taboo + RSA, Backbone=Llama-3-8B-Instruct2026.04 | 78.78 | 73.83 | 57.17 | 93.4 | 10.35 | 21.49 | 46.19 | 54.46 | |
| PPOTraining Set=Taboo + RSA, Backbone=Llama-3-8B-Instruct2026.04 | 78.29 | 72 | 60.99 | 97.8 | 12.51 | 25.8 | 49.58 | 56.71 | |
| ICTTraining Set=Taboo + RSA, Backbone=Llama-3-8B-Instruct2026.04 | 77.13 | 71.13 | 62.81 | 91.8 | 11.96 | 22.94 | 46.16 | 54.85 |