Chart Question Answering on ChartX (test)
92.99Annotated AccuracyGPT 4.1
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GPT 4.1Prompting Style=Highest of Zero-shot/CoT2025.10 | 92.99 | 69.58 | 77.9 | 80.25 | 78.56 | |
| GPT o3Prompting Style=Highest of Zero-shot/CoT2025.10 | 91.18 | 71.13 | 79.59 | 76.85 | 78.82 | |
| GPT o4-miniPrompting Style=Highest of Zero-shot/CoT2025.10 | 91.18 | 72.68 | 80.92 | 76.85 | 79.77 | |
| Gemini 2.0 flashPrompting Style=Highest of Zero-shot/CoT2025.10 | 89.37 | 58.31 | 68.72 | 74.07 | 70.23 | |
| Claude 3.7 SonnetPrompting Style=Highest of Zero-shot/CoT2025.10 | 89.37 | 60.28 | 69.81 | 75.62 | 71.44 | |
| Claude 3.5 SonnetPrompting Style=Highest of Zero-shot/CoT2025.10 | 87.78 | 57.32 | 67.39 | 73.15 | 69.01 | |
| MistralPrompting Style=Highest of Zero-shot/CoT2025.10 | 84.84 | 48.59 | 59.06 | 71.3 | 62.5 | |
| ChartAgentPrompting Style=Highest of Zero-shot/CoT, Framework=Multimodal Agentic Framework2025.10 | 84.84 | 44.16 | 55.93 | 69.14 | 59.69 | |
| Claude 3.5 HaikuPrompting Style=Highest of Zero-shot/CoT2025.10 | 80.32 | 40.7 | 50.97 | 68.52 | 55.9 |