Skip to content
/ GAPX Public

[NeurIPS 2022]GAPX: Generalized Autoregressive Paraphrase-identification X

Notifications You must be signed in to change notification settings

CUAI/GAPX

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

11 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

generalized_paraphrase_identification

Implementation of the paper 'GAPX: Generalized Autoregressive Paraphrase-identification X'

NeurIPS 2022

An ensemble model for paraphrase identification robust to distribution shift.

Canvas 1

Requirements

  • GPU
  • requirements.txt

Dataset

Please download the following paraphrase identification datasets:

Usage

To train and evaluate a paraphrase identification model, run:

python run.py --source_dataset [QQP, PIT, PAWS] --option [naive, robust]

Here we implemented a simplified version from the paper, where for the discriminative model, we use BART instead of RoBERTa

Results

You should expect to see something similar to this (f1/acc/auc):

Command QQP->QQP QQP->WMT QQP->PAWS QQP->PIT
python run.py --source_dataset QQP --option naive 83.4/83.5/91.2 66.7/66.8/74.2 44.7/49.8/57.1 63.6/66.5/82.0
python run.py --source_dataset QQP --option robust 83.1/83.2/88.4 74.4/74.7/79.3 56.6/56.9/59.5 62.3/63.6/73.5

About

[NeurIPS 2022]GAPX: Generalized Autoregressive Paraphrase-identification X

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages