Image2Description on Synthetic Eval Dataset Flowchart
0.9914BLEULlama-VL-TUG
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Llama-VL-TUG2026.01 | 0.9914 | 0.9914 | 0.9993 | 0.9996 | 0.8069 | 1 | |
| Qwen2.5-VL-7B-Instruct2026.01 | 0.6823 | 0.6823 | 0.8161 | 0.8432 | 0.4034 | 0.8936 | |
| Llama3.2-11B-Vision-Instruct2026.01 | 0.6598 | 0.6598 | 0.7943 | 0.8253 | 0.3285 | 0.845 | |
| Gemma3-12B-Instruction-Tuned2026.01 | 0.4695 | 0.4695 | 0.6673 | 0.7312 | 0.2282 | 0.9125 | |
| MiniCPM-V-2-62026.01 | 0.297 | 0.297 | 0.5195 | 0.4773 | -0.9542 | 0.4732 | |
| GPT-4o-mini2026.01 | 0.2728 | 0.2728 | 0.5799 | 0.5 | -0.977 | 0.4993 |