Dear author, I just attempted to reproduce your results using the parameters provided in Paper A.3, and the performance on the MMVP dataset is A little over 10 points lower (43% vs 56.8% in your paper). The screenshot of the running result is as follows. Could you please provide the specific running script for each dataset? Thank you!
