README.md

READ

This repository is the official implementation of our ACL'23 Findings paper Robust Natural Language Understanding with Residual Attention Debiasing.

Installation

Dependency

Experiments are run in the following environment:

Package	Version
conda	22.9.0
Python	3.8
CUDA	11.8

Install via Conda and Pip

conda create -n read python=3.8
conda activate read
pip install -r requirements.txt

Data

Download:

• FEVER train/dev/symmetric test data from FeverSymmetric GitHub repo
• For PAWS-QQP:
  • Generate PAWS-QQP data by following the instructions in the README file of the PAWS: Paraphrase Adversaries from Word Scrambling repo
  • Put the generated train.tsv and dev_and_test.tsv under qqp_paws directory The dataset directory should contain the following files:

└── dataset 
    └── fever
        ├── fever.dev.jsonl
        ├── fever.train.jsonl
        └── fever_symmetric_generated.jsonl
    └── qqp_paws
        ├── dev_and_test.tsv
        └── train.tsv

Training

Parameters are defined in train_ensemble.sh script. Change the value of task_name to the desired task name (mnli, fever, qqp). To train the ensemble model from scratch, please run the following

bash train_ensemble.sh

Evaluation

Due to different number of labels that MNLI and HANS datasets have, please:

• Go to eval_hans.sh
• Change value of model_name_or_path to your checkpoint
• Run the following to evaluate model on HANS dataset:

  bash eval_hans.sh
  

Citation

@inproceedings{wang-etal-2023-robust,
    title = "Robust Natural Language Understanding with Residual Attention Debiasing",
    author = "Wang, Fei  and  Huang, James Y.  and  Yan, Tianyi  and  Zhou, Wenxuan  and  Chen, Muhao",
    booktitle = "Findings of the Association for Computational Linguistics: ACL 2023",
    month = jul,
    year = "2023",
    address = "Toronto, Canada",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2023.findings-acl.32",
    doi = "10.18653/v1/2023.findings-acl.32",
    pages = "504--519",
}