Instruction Following on Average (AlpacaEval, Evol-Inst, UltraFeed)
90.9Average Win RateTeacher
Evaluation Results
| Method | Links | |
|---|---|---|
| TeacherModel=Qwen2.5-14B-Instruct2025.10 | 90.9 | |
| DistiLLM-2α=-5, Teacher Model=Qwen2.5-14B-Instruct, Student Model=Qwen2.5-1.5B-Instruct2025.10 | 72 | |
| DistiLLM-2α=-1, Teacher Model=Qwen2.5-14B-Instruct, Student Model=Qwen2.5-1.5B-Instruct2025.10 | 70.8 | |
| StudentModel=Qwen2.5-1.5B-Instruct2025.10 | 50.6 |