Test set size mismatch - Talk2Car dataset has 2,446 samples but evaluation expects 3,610

Small Dataset
#1

Hi everyone,

I’m trying to submit predictions for the Talk2Car task but encountering a size mismatch:

  • My predictions: 2,439 samples
  • AICrowd expects: 3,610 samples
  • Latest Talk2Car test set: 2,446 samples

Questions:

  1. Which version of Talk2Car should we use?
  2. Where can I find the test set with 3,610 samples?
  3. Is this an extended version or different split?

Has anyone else encountered this issue? Any guidance would be appreciated!

Thanks!