General Language Understanding on CMMLU
77.3Overall AccuracyNBDiff-7B-BASE
Evaluation results
| Method | Links | |
|---|---|---|
| NBDiff-7B-BASEConfiguration=Base2025.12 | 77.3 | |
| LLaDA-8BConfiguration=Base2025.12 | 69.9 | |
| LLaDA-MoE-7BConfiguration=A1B-Base2025.12 | 65.7 | |
| Dream-v0Configuration=Base-7B2025.12 | 60.9 | |
| InfLLM v2Training Tokens=100B, Inference Top-K=N/A, Training Top-K=22026.01 | 55.9 | |
| InfLLM v2Training Tokens=100B, Inference Top-K=2, Training Top-K=22026.01 | 55.53 | |
| InfLLM v2Training Tokens=100B, Inference Top-K=4, Training Top-K=42026.01 | 55.51 | |
| InfLLM v2Training Tokens=100B, Inference Top-K=N/A, Training Top-K=42026.01 | 55.48 | |
| DenseTraining Tokens=100B, Inference Top-K=22026.01 | 55.44 | |
| DenseTraining Tokens=100B, Inference Top-K=N/A2026.01 | 55.38 | |
| PHSATraining Tokens=100B, Inference Top-K=N/A, Training Top-K=42026.01 | 55.23 | |
| PHSATraining Tokens=100B, Inference Top-K=N/A, Training Top-K=22026.01 | 55.16 | |
| PHSATraining Tokens=100B, Inference Top-K=2, Training Top-K=22026.01 | 55.16 | |
| PHSATraining Tokens=100B, Inference Top-K=4, Training Top-K=42026.01 | 54.96 |