Chart-to-code Generation on ChartMIMIC 1.0 (test)
86.31Text F1 ScoreMETAL
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| METALBase Model=GPT-4O, n=52025.02 | 86.31 | 84.17 | 79.86 | 95.5 | 86.46 | |
| Hint-Enhanced PromptingBase Model=GPT-4O2025.02 | 77.02 | 80.84 | 72.75 | 93.89 | 81.12 | |
| Best-of-NBase Model=GPT-4O, n=52025.02 | 75.47 | 82.16 | 75.3 | 96.37 | 82.32 | |
| Direct PromptingBase Model=GPT-4O2025.02 | 74.83 | 81.24 | 74.24 | 94.76 | 81.26 | |
| METALBase Model=LLAMA 3.2-11B, n=52025.02 | 46.69 | 54.42 | 47.32 | 58.69 | 51.78 | |
| Best-of-NBase Model=LLAMA 3.2-11B, n=52025.02 | 40.28 | 36.6 | 38.43 | 57.22 | 43.13 | |
| Hint-Enhanced PromptingBase Model=LLAMA 3.2-11B2025.02 | 38.82 | 38.47 | 36.82 | 51.22 | 41.33 | |
| Direct PromptingBase Model=LLAMA 3.2-11B2025.02 | 36.7 | 37.07 | 33.46 | 54.56 | 40.45 |