Image2Code Generation on C4 Synthetic Eval Dataset
0.9585BLEULlama-VL-TUG
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Llama-VL-TUG2026.01 | 0.9585 | 95.8483 | 0.9845 | 98.9167 | 0.2863 | 0.9877 | |
| GPT-4o-mini2026.01 | 0.1022 | 10.4744 | 0.2325 | 44.4395 | -0.4793 | 0.4898 | |
| Qwen2.5-VL-7B-Instruct2026.01 | 0.0657 | 6.9249 | 0.1978 | 39.0591 | -0.5121 | 0.3557 | |
| Gemma3-12B-Instruction-Tuned2026.01 | 0.057 | 6.2295 | 0.2171 | 42.5174 | -0.4602 | 0.4325 | |
| Llama3.2-11B-Vision-Instruct2026.01 | 0.0525 | 5.5587 | 0.1569 | 37.4066 | -0.4687 | 0.358 | |
| MiniCPM-V-2-62026.01 | 0 | 0.0322 | 0.0073 | 1.8335 | -1.3834 | 0.0298 |