# Towards Robustifying NLI Models Against Lexical Dataset Biases

arXiv: https://arxiv.org/abs/2005.04732 · HTML: https://mlsanity.com/paper/2005.04732/

## Key facts

- Cited by 34 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 10 May 2020.

## Built on by (most cited first, 30 of 34)

- [Shortcut Learning of Large Language Models in Natural Language Understanding](https://mlsanity.com/paper/2208.11857/) (2022)
- [Towards Interpreting and Mitigating Shortcut Learning Behavior of NLU Models](https://mlsanity.com/paper/2103.06922/) (2021)
- [Identifying and Mitigating Spurious Correlations for Improving Robustness in NLP Models](https://mlsanity.com/paper/2110.07736/) (2021)
- [Generalization in NLI: Ways (Not) To Go Beyond Simple Heuristics](https://mlsanity.com/paper/2110.01518/) (2021)
- [Generating Data to Mitigate Spurious Correlations in Natural Language Inference Datasets](https://mlsanity.com/paper/2203.12942/) (2022)
- [Explore Spurious Correlations at the Concept Level in Language Models for Text Classification](https://mlsanity.com/paper/2311.08648/) (2023)
- [Flexible Generation of Natural Language Deductions](https://mlsanity.com/paper/2104.08825/) (2021)
- [Evaluating Gender Bias in Natural Language Inference](https://mlsanity.com/paper/2105.05541/) (2021)
- [MENLI: Robust Evaluation Metrics from Natural Language Inference](https://mlsanity.com/paper/2208.07316/) (2022)
- [Can Question Generation Debias Question Answering Models? A Case Study on Question-Context Lexical Overlap](https://mlsanity.com/paper/2109.11256/) (2021)
- [PECO: Examining Single Sentence Label Leakage in Natural Language Inference Datasets through Progressive Evaluation of Cluster Outliers](https://mlsanity.com/paper/2112.09237/) (2021)
- [Shortcut Learning in In-Context Learning: A Survey](https://mlsanity.com/paper/2411.02018/) (2024)
- [NLI Data Sanity Check: Assessing the Effect of Data Corruption on Model Performance](https://mlsanity.com/paper/2104.04751/) (2021)
- [DialGuide: Aligning Dialogue Model Behavior with Developer Guidelines](https://mlsanity.com/paper/2212.10557/) (2022)
- [SETI: Systematicity Evaluation of Textual Inference](https://mlsanity.com/paper/2305.15045/) (2023)
- [Learning Shortcuts: On the Misleading Promise of NLU in Language Models](https://mlsanity.com/paper/2401.09615/) (2024)
- [Resolving Lexical Bias in Model Editing](https://mlsanity.com/paper/2408.10411/) (2024)
- [Degendering Resumes for Fair Algorithmic Resume Screening](https://mlsanity.com/paper/2112.08910/) (2021)
- [Uncovering More Shallow Heuristics: Probing the Natural Language Inference Capacities of Transformer-Based Pre-Trained Language Models Using Syllogistic Patterns](https://mlsanity.com/paper/2201.07614/) (2022)
- [Enhancing Tabular Reasoning with Pattern Exploiting Training](https://mlsanity.com/paper/2210.12259/) (2022)
- [Looking at the Overlooked: An Analysis on the Word-Overlap Bias in Natural Language Inference](https://mlsanity.com/paper/2211.03862/) (2022)
- [HonestBait: Forward References for Attractive but Faithful Headline Generation](https://mlsanity.com/paper/2306.14828/) (2023)
- [SMoA: Sparse Mixture of Adapters to Mitigate Multiple Dataset Biases](https://mlsanity.com/paper/2302.14413/) (2023)
- [Designing NLP Systems That Adapt to Diverse Worldviews](https://mlsanity.com/paper/2405.11197/) (2024)
- [Unveiling Project-Specific Bias in Neural Code Models](https://mlsanity.com/paper/2201.07381/) (2022)
- [Multi-Scales Data Augmentation Approach In Natural Language Inference For Artifacts Mitigation And Pre-Trained Model Optimization](https://mlsanity.com/paper/2212.08756/) (2022)
- [Target-Aware Contextual Political Bias Detection in News](https://mlsanity.com/paper/2310.01138/) (2023)
- [On the Persistent Effects of Lexicality in Large Language Models](https://mlsanity.com/paper/2606.02750/) (2026)
- [An Understanding-Oriented Robust Machine Reading Comprehension Model](https://mlsanity.com/paper/2207.00187/) (2022)
- [Augmenting NLP data to counter Annotation Artifacts for NLI Tasks](https://mlsanity.com/paper/2302.04700/) (2023)

## Abstract (arXiv)

While deep learning models are making fast progress on the task of Natural Language Inference, recent studies have also shown that these models achieve high accuracy by exploiting several dataset biases, and without deep understanding of the language semantics. Using contradiction-word bias and word-overlapping bias as our two bias examples, this paper explores both data-level and model-level debiasing methods to robustify models against lexical dataset biases. First, we debias the dataset through data augmentation and enhancement, but show that the model bias cannot be fully removed via this method. Next, we also compare two ways of directly debiasing the model without knowing what the dataset biases are in advance. The first approach aims to remove the label bias at the embedding level. The second approach employs a bag-of-words sub-model to capture the features that are likely to exploit the bias and prevents the original model from learning these biased features by forcing orthogonality between these two sub-models. We performed evaluations on new balanced datasets extracted from the original MNLI dataset as well as the NLI stress tests, and show that the orthogonality approach is better at debiasing the model while maintaining competitive overall accuracy. Our code and data are available at: https://github.com/owenzx/LexicalDebias-ACL2020