General Knowledge (STEM) on MMLU STEM
74.1AccuracyBase Model
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Base ModelBackbone=Qwen3-1.7B2026.02 | 74.1 | 1,081 | — | |
| Extra-CoTBackbone=Qwen3-1.7B, Ratio=0.82026.02 | 71.3 | 1,267 | 1.17 | |
| Extra-CoTBackbone=Qwen3-1.7B, Ratio=1.02026.02 | 70.5 | 1,252 | 1.16 | |
| Extra-CoTBackbone=Qwen3-1.7B, Ratio=<POLICY>2026.02 | 67.5 | 1,026 | 0.95 | |
| Extra-CoT (CHRPO)Backbone=Qwen3-1.7B, Ratio=<POLICY>2026.02 | 67.1 | 330 | 0.31 | |
| Extra-CoTBackbone=Qwen3-1.7B, Ratio=0.42026.02 | 61.3 | 730 | 0.67 | |
| Extra-CoTBackbone=Qwen3-1.7B, Ratio=0.62026.02 | 60.6 | 1,012 | 0.94 | |
| TokenSkipBackbone=Qwen3-1.7B, Ratio=1.02026.02 | 56 | 1,556 | 1.44 | |
| TokenSkipBackbone=Qwen3-1.7B, Ratio=0.82026.02 | 54.5 | 1,481 | 1.37 | |
| Extra-CoTBackbone=Qwen3-1.7B, Ratio=0.22026.02 | 54.3 | 458 | 0.42 | |
| Gemma2-9BRole=Teacher2026.02 | 50.2 | — | — | |
| TokenSkipBackbone=Qwen3-1.7B, Ratio=0.62026.02 | 47.5 | 1,310 | 1.21 | |
| Qwen3-8B (teacher)Role=Teacher2026.02 | 40.2 | — | — | |
| TSD-KDTeacher=Qwen3-8B, Student=Qwen3-1.7B2026.02 | 37.5 | — | — | |
| Sequence KDTeacher=Qwen3-8B, Student=Qwen3-1.7B2026.02 | 36.9 | — | — | |
| DistiLLMTeacher=Qwen3-8B, Student=Qwen3-1.7B2026.02 | 36.8 | — | — | |
| GKDTeacher=Qwen3-8B, Student=Qwen3-1.7B2026.02 | 36.7 | — | — | |
| Supervised KDTeacher=Qwen3-8B, Student=Qwen3-1.7B2026.02 | 36.5 | — | — | |
| MiniLLMTeacher=Qwen3-8B, Student=Qwen3-1.7B2026.02 | 36.5 | — | — | |
| Qwen3-1.7B (student)Role=Student2026.02 | 36.2 | — | — | |
| TokenSkipBackbone=Qwen3-1.7B, Ratio=<POLICY>2026.02 | 32.6 | 1,035 | 0.96 | |
| TokenSkipBackbone=Qwen3-1.7B, Ratio=0.42026.02 | 29.8 | 877 | 0.81 | |
| TSD-KD2026.02 | 27 | — | — | |
| GKDβ=0.92026.02 | 26.2 | — | — | |
| Sequence-Level KD2026.02 | 25.9 | — | — | |
| MiniLLM2026.02 | 25.9 | — | — | |
| Gemma2-2BRole=Student2026.02 | 25.5 | — | — | |
| Supervised-KD2026.02 | 25.5 | — | — | |
| DistiLLM2026.02 | 25.2 | — | — | |
| TokenSkipBackbone=Qwen3-1.7B, Ratio=0.22026.02 | 18.1 | 446 | 0.41 |