Image2Code Generation on Synthetic Eval Dataset Sequence
82.36BLEULlama-VL-TUG
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Llama-VL-TUG2026.01 | 82.36 | 82.3598 | 95.55 | 96.2541 | 0.0992 | 96.06 | |
| Qwen2.5-VL-7B-Instruct2026.01 | 30.08 | 30.8903 | 63.68 | 73.6891 | 0.1591 | 60.02 | |
| Gemma3-12B-Instruction-Tuned2026.01 | 27.98 | 28.2498 | 64.16 | 73.1667 | 0.1889 | 58.6 | |
| Llama3.2-11B-Vision-Instruct2026.01 | 20.53 | 21.3545 | 56.46 | 65.3644 | -0.165 | 48.53 | |
| GPT-4o-mini2026.01 | 14.48 | 18.9303 | 58.75 | 67.1868 | -0.2169 | 48.63 | |
| MiniCPM-V-2-62026.01 | 0 | 1.7125 | 19.46 | 23.0416 | -1.3419 | 6.96 |