Mathematical Reasoning on GSM8K (MV, EM, PK)
75.04Mean Value (MV)I-PPO
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| I-PPOBackbone=Gemma2026.04 | 75.04 | 68.41 | 85.23 | |
| I-PPOBackbone=LLaMa2026.04 | 74.83 | 68.26 | 90.35 | |
| PPOBackbone=Gemma2026.04 | 72.94 | 63.98 | 84.39 | |
| PPOBackbone=LLaMa2026.04 | 70.86 | 56.14 | 88.39 | |
| GRPOBackbone=Gemma2026.04 | 70.73 | 62.5 | 83.02 | |
| BaselineBackbone=Gemma2026.04 | 70.13 | 61.95 | 83.02 | |
| GRPOBackbone=LLaMa2026.04 | 68.21 | 45.18 | 88.25 | |
| BaselineBackbone=LLaMa2026.04 | 65.58 | 34.99 | 88.25 | |
| I-PPOBackbone=Rho2026.04 | 51.93 | 46.1 | 74.83 | |
| PPOBackbone=Rho2026.04 | 50.05 | 41.11 | 73.09 | |
| GRPOBackbone=Rho2026.04 | 45.14 | 40.83 | 72.51 | |
| BaselineBackbone=Rho2026.04 | 43.52 | 40.26 | 72.48 |