Image2Description on Synthetic Eval Dataset C4
0.9916BLEULlama-VL-TUG
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Llama-VL-TUG2026.01 | 0.9916 | 99.1557 | 0.9993 | 99.9638 | 0.7393 | 1 | |
| GPT-4o-mini2026.01 | 0.8198 | 81.9774 | 0.9085 | 90.8801 | 0.5639 | 0.8511 | |
| Gemma3-12B-Instruction-Tuned2026.01 | 0.7599 | 75.9947 | 0.879 | 87.2885 | 0.477 | 0.8213 | |
| Qwen2.5-VL-7B-Instruct2026.01 | 0.7079 | 70.7854 | 0.834 | 84.7851 | 0.4182 | 0.8037 | |
| Llama3.2-11B-Vision-Instruct2026.01 | 0.5537 | 55.3658 | 0.7376 | 73.424 | 0.101 | 0.686 | |
| MiniCPM-V-2-62026.01 | 0.1583 | 15.8265 | 0.362 | 36.0967 | -0.4531 | 0.4934 |