Mathematical Reasoning on Comp. Math (val)
62.7AccuracyBase
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| BaseBackbone=Qwen3-32B, Generations=642026.04 | 62.7 | 3.67 | 34.7 | |
| ControlBackbone=Qwen3-32B, Generations=642026.04 | 58.5 | 4.51 | 48.5 | |
| MEMENTOBackbone=Qwen3-32B, Generations=642026.04 | 56.2 | 1.74 | 15.7 | |
| BaseBackbone=Phi-4-r (14B), Generations=642026.04 | 55.1 | 3.06 | 35.9 | |
| BaseBackbone=Qwen3-8B, Generations=642026.04 | 54.3 | 2.71 | 30.9 | |
| BaseBackbone=Olmo 3 (7B), Generations=82026.04 | 52.7 | 4.21 | 60.1 | |
| ControlBackbone=Phi-4-r (14B), Generations=642026.04 | 51.4 | 3.48 | 36.7 | |
| Mem. + RLBackbone=Qwen3-8B, Generations=642026.04 | 49.4 | 1.48 | 16.4 | |
| ControlBackbone=Qwen3-8B, Generations=642026.04 | 49.2 | 2.82 | 33.1 | |
| MEMENTOBackbone=Phi-4-r (14B), Generations=642026.04 | 48.7 | 1.25 | 13.1 | |
| ControlBackbone=Olmo 3 (7B), Generations=82026.04 | 48.3 | 3.78 | 46 | |
| MEMENTOBackbone=Olmo 3 (7B), Generations=82026.04 | 48.1 | 3.43 | 43.6 | |
| MEMENTOBackbone=Qwen3-8B, Generations=642026.04 | 45.1 | 1.08 | 10.7 |