The results of the plus experiment on the chartQA dataset could not be reproduced

[liuJP2]

Hi, I would like to know what prompt format to use in chartQA data set. However, I did not find a sample chartQA task template in the paper or readme file, can you let me know if it is convenient?

[liuJP2]

here is my prompt on chartQA dataset, by using it, I couldn't reproduce the results of the github report.

Here is the chart to answer this question. Please provide your explanation first, then answer the question in a short phrase starting by 'the answer is:'. Q: {}

In addition, I would like to ask how the final answer is extracted from the responses generated by the model?