This repo contains the few-shot data we used in our paper A Closer Look at Few-Shot Crosslingual Transfer: The Choice of Shots Matters. There are 40 sets of K-shot, K = 1, 2, 4 (,8) data sampled from each of the original dataset.
Note that for MLDoc we are not allowed to provide the actual corpus due to licence restrictions (cf. github.com/facebookresearch/MLDoc). We will provide the document indices soon.
The pickle format of the data is available at:
https://drive.google.com/file/d/1Vi1Cu9qx_2e44FLEHi1Z_Auh483j22ZD/view?usp=sharing