Mathematical Reasoning on MathVerse VO (Acc@1, Acc@16)
73.85Top-1 AccuracyQwen3-VL-8B-Instruct (Teacher)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen3-VL-8B-Instruct (Teacher)Model Size=8B2026.05 | 73.85 | — | |
| VGS On-Policy DistillationStudent Model=Qwen3-VL-4B-Instruct2026.05 | 71.1 | 68.53 | |
| Standard On-Policy DistillationStudent Model=Qwen3-VL-4B-Instruct2026.05 | 70.18 | 68.33 | |
| VGS On-Policy DistillationStudent Model=Qwen3-VL-2B-Instruct2026.05 | 56.19 | 56.52 | |
| Standard On-Policy DistillationStudent Model=Qwen3-VL-2B-Instruct2026.05 | 54.59 | 55.76 | |
| Qwen3-VL-4B-Instruct (Initial Student)Model Size=4B2026.05 | 53.67 | — | |
| Qwen3-VL-2B-Instruct (Initial Student)Model Size=2B2026.05 | 35.32 | — |