Instruction Following on UltraFeed
84.9Win RateTeacher
Evaluation Results
| Method | Links | |
|---|---|---|
| TeacherModel=Qwen2.5-14B-Instruct2025.10 | 84.9 | |
| MT (Teacher)Distillation Pair (Teacher Model -> Student Model)=Qwen2.5-7B-Inst -> Qwen2.5-1.5B2026.04 | 77.16 | |
| DistilLLMDistillation Pair (Teacher Model -> Student Model)=Qwen2.5-7B-Inst -> Qwen2.5-1.5B2026.04 | 65.92 | |
| VCRDDistillation Pair (Teacher Model -> Student Model)=Qwen2.5-7B-Inst -> Qwen2.5-1.5B2026.04 | 65.67 | |
| DistillLM-2Distillation Pair (Teacher Model -> Student Model)=Qwen2.5-7B-Inst -> Qwen2.5-1.5B2026.04 | 65.57 | |
| MT (Teacher)Distillation Pair (Teacher Model -> Student Model)=Qwen2-7B-Inst -> Qwen2-1.5B2026.04 | 64.86 | |
| DistiLLM-2α=-5, Teacher Model=Qwen2.5-14B-Instruct, Student Model=Qwen2.5-1.5B-Instruct2025.10 | 63.6 | |
| KDDistillation Pair (Teacher Model -> Student Model)=Qwen2.5-7B-Inst -> Qwen2.5-1.5B2026.04 | 62.6 | |
| DistiLLM-2α=-1, Teacher Model=Qwen2.5-14B-Instruct, Student Model=Qwen2.5-1.5B-Instruct2025.10 | 61.7 | |
| SeqKDDistillation Pair (Teacher Model -> Student Model)=Qwen2.5-7B-Inst -> Qwen2.5-1.5B2026.04 | 52.89 | |
| DistillLM-2Distillation Pair (Teacher Model -> Student Model)=Qwen2-7B-Inst -> Qwen2-1.5B2026.04 | 51.97 | |
| VCRDDistillation Pair (Teacher Model -> Student Model)=Qwen2-7B-Inst -> Qwen2-1.5B2026.04 | 51.56 | |
| DistilLLMDistillation Pair (Teacher Model -> Student Model)=Qwen2-7B-Inst -> Qwen2-1.5B2026.04 | 51.26 | |
| KDDistillation Pair (Teacher Model -> Student Model)=Qwen2-7B-Inst -> Qwen2-1.5B2026.04 | 49.25 | |
| MS (Student)Distillation Pair (Teacher Model -> Student Model)=Qwen2.5-7B-Inst -> Qwen2.5-1.5B2026.04 | 42.94 | |
| StudentModel=Qwen2.5-1.5B-Instruct2025.10 | 40.4 | |
| SeqKDDistillation Pair (Teacher Model -> Student Model)=Qwen2-7B-Inst -> Qwen2-1.5B2026.04 | 39.1 | |
| MS (Student)Distillation Pair (Teacher Model -> Student Model)=Qwen2-7B-Inst -> Qwen2-1.5B2026.04 | 38.2 |