Question of retrieving in-context examples.

[mtt1998]

Hi,
Thank you for sharing your code of this good work. I have a question for the prompt_retrieval function in the main file:
https://github.com/HKUNLP/icl-selective-annotation/blob/4f69831938f5fea6e6299efc55df2ecbe207fb60/main.py#L115

It seems that this function retrieves kNN examples from the "train_examples" instead of "processed_train_examples".
And the "train_examples" contains all examples in the train dataset, the "processed_train_examples" contains all selected examples for annotation. Is this a bug or I misunderstand anything? Could you help me with this question? Thank you~