This repository contains the frontier research on counterfactual generation which is a hot topic recently. A counterfactual is obtained by making minimal or loose changes to existing data to flip their label for a particular NLP task. The counterfactually augmented data (CAD) consists of original data and counterfactuals, also called contrast sets (Gardner et al., 2020; Atanasova et al., 2022).
Here are some examples of counterfactuals based on sentiment analysis and natural language inference tasks.
The following papers are collected and categorized according to our survey paper—— A Survey on Natural Language Counterfactual Generation.
Counterfactual Generation Taxonomy:
Matt Gardner, et al. Evaluating Models’ Local Decision Boundaries via Contrast Sets. In Findings of the Association for Computational Linguistics: EMNLP 2020, pages 1307–1323, Online. Association for Computational Linguistics.
Pepa Atanasova, et al. Fact Checking with Insufficient Evidence. Transactions of the Association for Computational Linguistics (TACL), 2022.
Counterfactual story reasoning and generation, EMNLP-IJCNLP 2019
Learning the difference that makes a difference with counterfactually-augmented data, ICLR 2020
More bang for your buck: Natural perturbation for robust question answering, EMNLP 2020
Evaluating Models’ Local Decision Boundaries via Contrast Sets, EMNLP finding 2020
Automated Fact-Checking of Claims from Wikipedia, LREC 2020
Generate Your Counterfactuals: Towards Controlled Counterfactual Generation for Text, AAAI 2021
Generating Realistic Natural Language Counterfactuals, EMNLP finding 2021
A causal lens for controllable text generation, NeurIPS 2021
Counterfactual Explanation Based on Gradual Construction for Deep Networks, Pattern Recognition 2022
Counterfactual Sentence Generation with Plug-and-Play Perturbation, SaTML 2023
Counterfactual generation with identifiability guarantees, NeurIPS 2023
Explaining Data-Driven Document Classifications, MIS Quarterly 2014
Delete, Retrieve, Generate: A Simple Approach to Sentiment and Style Transfer, NAACL 2018
Mask and Infill: Applying Masked Language Model for Sentiment Transfer, IJCAI 2019
PAWS: Paraphrase Adversaries from Word Scrambling, NAACL-HLT 2019
Politeness transfer: A tag and generate approach, ACL 2020
Unsupervised Text Style Transfer with Padded Masked Language Models, EMNLP 2020
Linguistically-informed transformations (LIT): A method for automatically generating contrast sets ACL 2020
Beyond Accuracy: Behavioral Testing of NLP Models with CheckList 2020 ACL
Generating Plausible Counterfactual Explanations for Deep Transformers in Financial Text Classification, Coling (CCFB) 2020
Logic-guided data augmentation and regularization for consistent question answering, ACL 2020.
LEWIS: Levenshtein editing for unsupervised text style transfer, ACL 2021
Polyjuice: Generating Counterfactuals for Explaining, Evaluating, and Improving Models, ACL 2021
Explaining NLP Models via Minimal Contrastive Editing (MiCE), ACL finding 2021
CrossAug: A Contrastive Data Augmentation Method for Debiasing Fact Verification Models, CIKM 2021
Exploring the Efficacy of Automatically Generated Counterfactuals for Sentiment Analysis, ACL | IJCNLP 2021
Entity-based knowledge conflicts in question answering, EMNLP 2021
Controlling Text Edition by Changing Answers of Specific Questions, ACL-IJCNLP finding 2021
KACE: Generating Knowledge Aware Contrastive Explanations for Natural Language Inference, ACL 2021
Reinforced Counterfactual Data Augmentation for Dual Sentiment Classification, EMNLP 2021
Learning Models for Actionable Recourse, NeurIPS 2021
Text Counterfactuals via Latent Optimization and Shapley-Guided Search, EMNLP 2021
Sketch and Customize: A Counterfactual Story Generator, AAAI 2021
Tailor: Generating and Perturbing Text with Semantic Controls, ACL 2022
Retrieval-guided Counterfactual Generation for QA, ACL 2022
Break, Perturb, Build: Automatic Perturbation of Reasoning Paths Through Question Decomposition, TACL 2022
AutoCAD: Automatically Generate Counterfactuals for Mitigating Shortcut Learning, EMNLP finding 2022
Let the CAT out of the bag: Contrastive Attributed explanations for Text, EMNLP 2022
NeuroCounterfactuals: Beyond Minimal-Edit Counterfactuals for Richer Data Augmentation, EMNLP finding 2022
DoCoGen: Domain Counterfactual Generation for Low Resource Domain Adaptation, ACL 2022
Unsupervised Editing for Counterfactual Stories, AAAI 2022
Relevance-based Infilling for Natural Language Counterfactuals CKLM 2023
CREST: A Joint Framework for Rationalization and Counterfactual Text Generation, ACL 2023
Towards Model Robustness: Generating Contextual Counterfactuals for Entities in Relation Extraction, WWW 2023
SCENE: Self-Labeled Counterfactuals for Extrapolating to Negative Examples, EMNLP 2023
Generating Commonsense Counterfactuals for Stable Relation Extraction, EMNLP 2023
CLICK: Integrating Causal Inference and Commonsense Knowledge Incorporation for Counterfactual Story Generation, Electronics 2023
A general search-based framework for generating textual counterfactual explanations, AAAI 2024
A Novel Counterfactual Data Augmentation Method for Aspect-Based Sentiment Analysis, ACML 2023
Counterfactual-Enhanced Information Bottleneck for Aspect-Based Sentiment Analysis, AAAI 2024
CORE: A Retrieve-then-Edit Framework for Counterfactual Data Generation, EMNLP finding 2022
DISCO: Distilling Counterfactuals with Large Language Models, ACL 2023
CATfOOD: Counterfactual Augmented Training for Improving Out-of-Domain Performance and Calibration, EACL 2024
Faithful Explanations of Black-box NLP Models Using LLM-generated Counterfactuals, ICLR Poster 2024
LLMs for Generating and Evaluating Counterfactuals: A Comprehensive Study, arXiv preprint 2024
Prompting Large Language Models for Counterfactual Generation: An Empirical Study, LREC-COLING 2024
Towards LLM-guided Causal Explainability for Black-box Text Classifiers, AAAI Workshop 2024
Explore Spurious Correlations at the Concept Level in Language Models for Text Classification, ACL 2024
Improving Classifier Robustness through Active Generation of Pairwise Counterfactuals, EMNLP finding 2023
