Explanation quality evaluation on Synthetic (test)
87.6HelpfulnessQwen3-VL-8b-SVR-FT
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen3-VL-8b-SVR-FTTraining=SVR-FT2025.12 | 87.6 | 83.9 | 95.3 | 89 | |
| GPT-5-miniModel Type=Base2025.12 | 73.4 | 65.1 | 79.2 | 72.6 | |
| Qwen3-VL-8b-GFTTraining=GFT2025.12 | 72.5 | 69.4 | 75.1 | 72.3 | |
| Qwen3-VL-8bModel Type=Base2025.12 | 58.2 | 45.2 | 75.2 | 59.5 | |
| Gemini-2.5-flashModel Type=Base2025.12 | 46.8 | 37.9 | 53.7 | 46.1 | |
| Qwen3-VL-8b-FTTraining=Fine-tuned2025.12 | 43.4 | 47.8 | 57.2 | 49.5 |