Mathematical Reasoning on MathVerse-VD
79.63Top-1 AccuracyQwen3-VL-8B-Instruct (Teacher)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen3-VL-8B-Instruct (Teacher)Model Size=8B2026.05 | 79.63 | — | |
| VGS On-Policy DistillationStudent Model=Qwen3-VL-4B-Instruct2026.05 | 74.31 | 73.89 | |
| Standard On-Policy DistillationStudent Model=Qwen3-VL-4B-Instruct2026.05 | 71.53 | 73.6 | |
| VGS On-Policy DistillationStudent Model=Qwen3-VL-2B-Instruct2026.05 | 58.1 | 60.1 | |
| Qwen3-VL-4B-Instruct (Initial Student)Model Size=4B2026.05 | 57.18 | — | |
| Standard On-Policy DistillationStudent Model=Qwen3-VL-2B-Instruct2026.05 | 56.02 | 59.68 | |
| Qwen3-VL-2B-Instruct (Initial Student)Model Size=2B2026.05 | 35.88 | — |