How to reproduce the Comet score in the CLASI paper.

Thank you for providing this dataset. I would like to know how to reproduce the COMET metric reported in the paper. I used the results from https://docs.google.com/spreadsheets/d/1OBDbQMlixqFkBfzLW5pmP8WsI-RHqAtYbLnkd321N6A/edit?gid=1328074313#gid=1328074313 and calculated the COMET score at the sentence level, but it only reached around 80, whereas the paper reports approximately 84. Are there any special processing steps required during the calculation?