Explanation quality evaluation on In-house Dataset
80.8HelpfulnessQwen3-VL-8b-SVR-FT
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Qwen3-VL-8b-SVR-FTTraining=SVR-FT2025.12 | 80.8 | 70.6 | 91.6 | 81 | 6.985 | |
| GPT-5-miniModel Type=Base2025.12 | 76.2 | 68.7 | 82.9 | 75.9 | 4.915 | |
| Qwen3-VL-8b-GFTTraining=GFT2025.12 | 71.3 | 67.9 | 74.2 | 71.1 | 6.445 | |
| Gemini-2.5-flashModel Type=Base2025.12 | 69.7 | 52.1 | 78.3 | 66.7 | 5.86 | |
| Qwen3-VL-8bModel Type=Base2025.12 | 56 | 36.8 | 75.5 | 56.1 | 2.44 | |
| Qwen3-VL-8b-FTTraining=Fine-tuned2025.12 | 40.1 | 37.1 | 59 | 45.4 | 2.178 |