Did you use clinc150 data_imbalanced vs data_full?

Hi,

I experimented with verifying your approach however I didn't receive such low F1 numbers for few shot. In looking at data_full, banking has the same amount of training samples as all other domains/intents. Did you in fact use data imbalanced?

Compare the reported CLINC150 experiment with the repository's data_full and data_imbalanced datasets, paying particular attention to the banking domain and training-sample counts. First identify which dataset the experiments used, then verify the few-shot F1 results. Done means documenting whether the discrepancy comes from dataset selection or another reproducible setup difference.

Written by the indexing model from the issue text.