Mathematical Reasoning on GSM8K (Expert Count, Accuracy)
71.5Expert CountLLaDA2.0-Mini
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LLaDA2.0-MiniZero-shot=true2026.05 | 71.5 | 90.8 | |
| Top-4Zero-shot=true2026.05 | 44.3 | 72.4 | |
| DES-SZero-shot=true2026.05 | 41.9 | 90.7 | |
| DES-VZero-shot=true2026.05 | 37 | 90.5 | |
| dMoEZero-shot=true, p_train=0.62026.05 | 14.7 | 88.6 | |
| DES-SZero-shot=true, Adjusted threshold=true2026.05 | 13.2 | 74 | |
| DES-VZero-shot=true, Adjusted threshold=true2026.05 | 12.9 | 73.5 | |
| dMoEZero-shot=true, p_train=0.52026.05 | 11.6 | 87.1 |