search

DFKI-NLP / InterroLang

InterroLang: Exploring NLP Models and Datasets through Dialogue-based Explanations [EMNLP 2023 Findings]
https://arxiv.org/abs/2310.05592
5 stars 1 forks source link

rationale precompute generator code #95

Closed schopra6 closed 1 year ago

schopra6 commented 1 year ago

json load operation added not been tested beta version

qiaw99 commented 1 year ago

The file for DA dataset is wrong, please use this one: https://github.com/nfelnlp/InterroLang/blob/main/data/da_test_set_with_indices.csv

qiaw99 commented 1 year ago

You should use this file https://github.com/nfelnlp/InterroLang/blob/main/data/offensive_val.csv for OLID. Please refer gin config file

nfelnlp commented 1 year ago

For BoolQ and DailyDialog, the formatting is not quite right yet. The "Explanation:" should only contain the output of the Dolly model.

BoolQ

Screenshot from 2023-06-05 10-41-29

DailyDialog

Screenshot from 2023-06-05 10-44-14