English Knowledge on MMLU-Redux
90.15Accuracy (English Knowledge)Teacher-30B
Evaluation Results
| Method | Links | |
|---|---|---|
| Teacher-30BInference Mode=thinking2026.06 | 90.15 | |
| Teacher-8BInference Mode=thinking2026.06 | 85.86 | |
| OPD-T30BInference Mode=thinking2026.06 | 84.89 | |
| OPD-T8BInference Mode=thinking2026.06 | 84.83 | |
| Qwen3.5-9BShots=52026.04 | 83.43 | |
| Llama-3.3-70B-InstructShots=52026.04 | 81.73 | |
| Qwen3-14BShots=52026.04 | 80.47 | |
| XekRung-8BShots=52026.04 | 78.77 | |
| Qwen3-8BShots=52026.04 | 77.53 | |
| RLInference Mode=thinking2026.06 | 74.82 | |
| Base (SFT ckpt)Inference Mode=thinking2026.06 | 69.75 | |
| SecGPT-14BShots=52026.04 | 69.5 | |
| Llama-Primus-Reasoning-8BShots=52026.04 | 21.8 |