The official implementation of our paper "TELLER: A Trustworthy Framework for Explainable, Generalizable and Controllable Fake News Detection".

Step 1: Download the dataset folder from onedrive by data.zip. Unzip this folder into the project directory. You can find four orginal datasets, pre-processed datasets (i.e., val.jsonl, test.jsonl, train.jsonl in each dataset folder) and the files incuding questions and answers

Step 2: Place you OpenAI key into the file named api_key.txt.

openai.api_key = ""

To reproduce the results of in-domain experiments on four Dataset:

python drive_liar.py

To reproduce the results of cross-domain experiments on three datasets:

python drive_dg.py