This repository contains the code for the following paper:

• Sicheng Yu, Yulei Niu, Shuohang Wang, Jing Jiang, Qianru Sun *"Counterfactual Variable Control for Robust and Interpretable Question Answering (https://arxiv.org/abs/2010.05581)

• torch 1.3.1
• transformers 2.1.1
• apex 0.1
• tensorboardX 1.8
• prettytable 0.7.2

Here we use RACE with BERT-base as example for MCQA task.

• Step 1: Download original dataset via this link (http://www.cs.cmu.edu/~glai1/data/race/), and store them in directory /data_mc/RACE.
• Step 2: Download the adversarial sets via this link (https://drive.google.com/drive/folders/1ufPl0aP-QglVdsDtlKq9kTnt_0Fqmw2i?usp=sharing), and store them in same directory as step 1.

cd src_mc
bash train.sh

You may visualize the loss trend using tensorboardX in directory /src_mc/runs.

Please change --timestamp according to your training time.

bash cvc_iv.sh

Please change --pre_model_dir according to model selected by you.

bash cvc_mv.sh

You can download the CVC model trained by us (CVC-MV is not included). You can find the results we reported in our paper. (https://drive.google.com/drive/folders/14ZMUwW_bxnpaDX4HbjdUxGwcNBzFIYR6?usp=sharing)

Coming Soon!