Chart-to-code generation on ChartMimic
0.86Text ScoreMETAL
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| METALBase Model=GPT-4o2025.06 | 0.86 | 0.84 | 0.96 | 0.8 | 13.09 | 0.7 | 11,356 | 6.89 | |
| ChartIRBase Model=GPT-4o2025.06 | 0.82 | 0.84 | 0.96 | 0.76 | 13.68 | 0.67 | 10,129 | 7.15 | |
| DirectBase Model=GPT-4o2025.06 | 0.78 | 0.82 | 0.97 | 0.74 | 13.05 | 0.64 | 11,846 | 6.62 | |
| ChartIRBase Model=Qwen2-VL2025.06 | 0.48 | 0.64 | 0.75 | 0.49 | 12.01 | 0.6 | 15,127 | 3.86 | |
| METALBase Model=Qwen2-VL2025.06 | 0.35 | 0.4 | 0.78 | 0.33 | 11.74 | 0.6 | 16,461 | 2.32 | |
| DirectBase Model=Qwen2-VL2025.06 | 0.3 | 0.41 | 0.78 | 0.41 | 11.73 | 0.6 | 16,644 | 2.2 |