Mathematical Reasoning on MATH 256
36.46AccuracyFull
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| FullBackbone=Dream-7B-Base, Sampling Strategy=Full-step2026.04 | 36.46 | 256 | 29.5 | |
| ThresholdBackbone=Dream-7B-Base, Sampling Strategy=Confidence Threshold2026.04 | 36.35 | 99.88 | 76.3 | |
| NI SamplingBackbone=Dream-7B-Base2026.04 | 35.74 | 72.58 | 100.4 | |
| FullModel=LLaDA 1.52026.04 | 33.18 | 256 | 25 | |
| ThresholdModel=LLaDA 1.52026.04 | 33.18 | 95.41 | 67 | |
| NI SamplingModel=LLaDA 1.52026.04 | 32.98 | 69.67 | 89.5 | |
| FullModel=LLaDA-8B-Instruct2026.04 | 31.89 | 256 | 25 | |
| ThresholdModel=LLaDA-8B-Instruct2026.04 | 31.68 | 95.98 | 66.7 | |
| NI SamplingModel=LLaDA-8B-Instruct2026.04 | 31.67 | 61.92 | 98.7 |