Mathematical Reasoning on GSM8K (Accuracy & NFE)
80.82AccuracyPRR
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PRRBackbone=LLaDA-8B2026.03 | 80.82 | 70.69 | |
| EB-SamplerBackbone=LLaDA-8B2026.03 | 80.74 | 74.72 | |
| VanillaBackbone=LLaDA-8B2026.03 | 80.21 | 256 | |
| Dynamic-SamplerBackbone=LLaDA-8B2026.03 | 79.23 | 73.23 | |
| PRRBackbone=Dream-7B2026.03 | 74.15 | 138.02 | |
| EB-SamplerBackbone=Dream-7B2026.03 | 73.69 | 141.5 | |
| VanillaBackbone=Dream-7B2026.03 | 73.62 | 256 | |
| Dynamic-SamplerBackbone=Dream-7B2026.03 | 72.93 | 138.68 |