Image2Description on Synthetic Eval Dataset Class
97.46BLEULlama-VL-TUG
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Llama-VL-TUG2026.01 | 97.46 | 97.4633 | 98.09 | 98.0992 | 0.6531 | 97.64 | |
| GPT-4o-mini2026.01 | 80.32 | 80.3173 | 83.33 | 87.2117 | 0.3869 | 68.36 | |
| Qwen2.5-VL-7B-Instruct2026.01 | 72.94 | 72.938 | 74.39 | 82.1311 | 0.3192 | 65.71 | |
| Gemma3-12B-Instruction-Tuned2026.01 | 69.78 | 69.7829 | 69.72 | 77.1858 | 0.2685 | 62.76 | |
| Llama3.2-11B-Vision-Instruct2026.01 | 32.56 | 32.5603 | 48.65 | 50.6936 | -0.2765 | 52.87 | |
| MiniCPM-V-2-62026.01 | 32.56 | 32.5603 | 48.65 | 50.6936 | -0.2765 | 52.87 |