Numbers mismatch for RSVQA-HR evaluation dataset?

[lx709]

Hello, authors,

Thank you for sharing the code for your excellent work. I'm currently conducting an evaluation on the HRBEN dataset and encountered a discrepancy. In your GeoChat paper, you mentioned that the Test Set-2 of RSVQA-HRBEN [20] contains 47k question-answer pairs. However, the hrben.json file on Hugging Face contains 62,554 question-answer pairs. Could you please clarify which number is correct?

Thank you for your attention to this matter.