# Generating Token-Level Explanations for Natural Language Inference

arXiv: https://arxiv.org/abs/1904.10717 · HTML: https://mlsanity.com/paper/1904.10717/

## Key facts

- Cited by 28 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 24 Apr 2019.

## Built on by (most cited first, 28 of 28)

- [Explainability for Large Language Models: A Survey](https://mlsanity.com/paper/2309.01029/) (2023)
- [Get Your Vitamin C! Robust Fact Verification with Contrastive Evidence](https://mlsanity.com/paper/2103.08541/) (2021)
- [Rethinking Attention-Model Explainability through Faithfulness Violation Test](https://mlsanity.com/paper/2201.12114/) (2022)
- [Supervising Model Attention with Human Explanations for Robust Natural Language Inference](https://mlsanity.com/paper/2104.08142/) (2021)
- [Connecting Attributions and QA Model Behavior on Realistic Counterfactuals](https://mlsanity.com/paper/2104.04515/) (2021)
- [Explainable Activity Recognition for Smart Home Systems](https://mlsanity.com/paper/2105.09787/) (2021)
- [Explaining Neural Network Predictions on Sentence Pairs via Learning Word-Group Masks](https://mlsanity.com/paper/2104.04488/) (2021)
- [Unifying Model Explainability and Robustness for Joint Text Classification and Rationale Extraction](https://mlsanity.com/paper/2112.10424/) (2021)
- [Attention Meets Post-hoc Interpretability: A Mathematical Perspective](https://mlsanity.com/paper/2402.03485/) (2024)
- [Interpretability Needs a New Paradigm](https://mlsanity.com/paper/2405.05386/) (2024)
- [Class-Discriminative Attention Maps for Vision Transformers](https://mlsanity.com/paper/2312.02364/) (2023)
- [Distinguish Before Answer: Generating Contrastive Explanation as Knowledge for Commonsense Question Answering](https://mlsanity.com/paper/2305.08135/) (2023)
- [Zero-shot Sequence Labeling for Transformer-based Sentence Classifiers](https://mlsanity.com/paper/2103.14465/) (2021)
- [Decoding Layer Saliency in Language Transformers](https://mlsanity.com/paper/2308.05219/) (2023)
- [Parallel Sentence-Level Explanation Generation for Real-World Low-Resource Scenarios](https://mlsanity.com/paper/2302.10707/) (2023)
- [You Can Do Better! If You Elaborate the Reason When Making Prediction](https://mlsanity.com/paper/2103.14919/) (2021)
- [Revisiting Attention Weights as Explanations from an Information Theoretic Perspective](https://mlsanity.com/paper/2211.07714/) (2022)
- [Multi-Layer Attention-Based Explainability via Transformers for Tabular Data](https://mlsanity.com/paper/2302.14278/) (2023)
- [MorphNLI: A Stepwise Approach to Natural Language Inference Using Text Morphing](https://mlsanity.com/paper/2502.09567/) (2025)
- [LIBERTy: A Causal Framework for Benchmarking Concept-Based Explanations of LLMs with Structural Counterfactuals](https://mlsanity.com/paper/2601.10700/) (2026)
- [Do Transformers know symbolic rules, and would we know if they did?](https://mlsanity.com/paper/2203.00162/) (2022)
- [Zero-shot Code-Mixed Offensive Span Identification through Rationale Extraction](https://mlsanity.com/paper/2205.06119/) (2022)
- [LMExplainer: Grounding Knowledge and Explaining Language Models](https://mlsanity.com/paper/2303.16537/) (2023)
- [Neural models for Factual Inconsistency Classification with Explanations](https://mlsanity.com/paper/2306.08872/) (2023)
- [Explainable Semantic Textual Similarity via Dissimilar Span Detection](https://mlsanity.com/paper/2603.21174/) (2026)
- [MSSRNet: Manipulating Sequential Style Representation for Unsupervised Text Style Transfer](https://mlsanity.com/paper/2306.07994/) (2023)
- [eXplainable Bayesian Multi-Perspective Generative Retrieval](https://mlsanity.com/paper/2402.02418/) (2024)
- [Regularizing Attention Scores with Bootstrapping](https://mlsanity.com/paper/2604.01339/) (2026)

## Abstract (arXiv)

The task of Natural Language Inference (NLI) is widely modeled as supervised sentence pair classification. While there has been a lot of work recently on generating explanations of the predictions of classifiers on a single piece of text, there have been no attempts to generate explanations of classifiers operating on pairs of sentences. In this paper, we show that it is possible to generate token-level explanations for NLI without the need for training data explicitly annotated for this purpose. We use a simple LSTM architecture and evaluate both LIME and Anchor explanations for this task. We compare these to a Multiple Instance Learning (MIL) method that uses thresholded attention make token-level predictions. The approach we present in this paper is a novel extension of zero-shot single-sentence tagging to sentence pairs for NLI. We conduct our experiments on the well-studied SNLI dataset that was recently augmented with manually annotation of the tokens that explain the entailment relation. We find that our white-box MIL-based method, while orders of magnitude faster, does not reach the same accuracy as the black-box methods.