Image2Code on Real-world images corpus Packet D3
4Code Error Rate (CEr)Llama-VL-TUG
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Llama-VL-TUGstatus=Fine-tuned, backbone=Llama3.2-11B-Vision-Instruct2026.01 | 4 | 40 | — | 15 | |
| Gemma3-12B2026.01 | 23 | 0 | — | 0 | |
| GPT-4o-mini2026.01 | 25 | 0 | — | 0 | |
| Qwen2.5-VL-7B2026.01 | 51 | 0 | — | 0 | |
| MiniCPM-V-2-62026.01 | 51 | 0 | — | 0 | |
| Llama3.2-11Btstatus=Vanilla2026.01 | 66 | 0 | — | 0 |