Image2Description on Synthetic Eval Dataset Sequence
0.8888BLEULlama-VL-TUG
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Llama-VL-TUG2026.01 | 0.8888 | 88.8756 | 0.9775 | 97.2891 | 0.618 | 0.9603 | |
| GPT-4o-mini2026.01 | 0.7568 | 75.6781 | 0.9242 | 92.3463 | 0.546 | 0.8144 | |
| Llama3.2-11B-Vision-Instruct2026.01 | 0.6893 | 68.9273 | 0.878 | 84.3601 | 0.3644 | 0.7752 | |
| MiniCPM-V-2-62026.01 | 0.6893 | 68.9273 | 0.878 | 84.3601 | 0.3644 | 0.7752 | |
| Qwen2.5-VL-7B-Instruct2026.01 | 0.6505 | 65.0489 | 0.8374 | 82.4184 | 0.3669 | 0.7782 | |
| Gemma3-12B-Instruction-Tuned2026.01 | 0.5279 | 52.7857 | 0.7633 | 79.9958 | 0.2875 | 0.7664 |