# Revisiting Attention Weights as Explanations from an Information Theoretic Perspective

arXiv: https://arxiv.org/abs/2211.07714 · HTML: https://mlsanity.com/paper/2211.07714/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 17 of its 25 references that are in the map.
- First posted on arXiv 31 Oct 2022.

## Builds on (most cited first, 17 of 17)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Adam: A Method for Stochastic Optimization](https://mlsanity.com/paper/1412.6980/) (2014)
- [Neural Machine Translation by Jointly Learning to Align and Translate](https://mlsanity.com/paper/1409.0473/) (2014)
- [Categorical Reparameterization with Gumbel-Softmax](https://mlsanity.com/paper/1611.01144/) (2016)
- [A large annotated corpus for learning natural language inference](https://mlsanity.com/paper/1508.05326/) (2015)
- [Show, Attend and Tell: Neural Image Caption Generation with Visual Attention](https://mlsanity.com/paper/1502.03044/) (2015)
- [Attention is not Explanation](https://mlsanity.com/paper/1902.10186/) (2019)
- [Deep Learning and the Information Bottleneck Principle](https://mlsanity.com/paper/1503.02406/) (2015)
- [Attention is not not Explanation](https://mlsanity.com/paper/1908.04626/) (2019)
- [Assessing Generative Models via Precision and Recall](https://mlsanity.com/paper/1806.00035/) (2018)
- [RETAIN: An Interpretable Predictive Model for Healthcare using Reverse Time Attention Mechanism](https://mlsanity.com/paper/1608.05745/) (2016)
- [From Softmax to Sparsemax: A Sparse Model of Attention and Multi-Label Classification](https://mlsanity.com/paper/1602.02068/) (2016)
- [Towards AI-Complete Question Answering: A Set of Prerequisite Toy Tasks](https://mlsanity.com/paper/1502.05698/) (2015)
- [Explainable Prediction of Medical Codes from Clinical Text](https://mlsanity.com/paper/1802.05695/) (2018)
- [The Bottom-up Evolution of Representations in the Transformer: A Study with Machine Translation and Language Modeling Objectives](https://mlsanity.com/paper/1909.01380/) (2019)
- [Explaining a black-box using Deep Variational Information Bottleneck Approach](https://mlsanity.com/paper/1902.06918/) (2019)
- [Generating Token-Level Explanations for Natural Language Inference](https://mlsanity.com/paper/1904.10717/) (2019)

## Built on by (most cited first, 2 of 2)

- [Semantic Interpretation and Validation of Graph Attention-based Explanations for GNN Models](https://mlsanity.com/paper/2308.04220/) (2023)
- [Introspection in Learned Semantic Scene Graph Localisation](https://mlsanity.com/paper/2510.07053/) (2025)

## Abstract (arXiv)

Attention mechanisms have recently demonstrated impressive performance on a range of NLP tasks, and attention scores are often used as a proxy for model explainability. However, there is a debate on whether attention weights can, in fact, be used to identify the most important inputs to a model. We approach this question from an information theoretic perspective by measuring the mutual information between the model output and the hidden states. From extensive experiments, we draw the following conclusions: (i) Additive and Deep attention mechanisms are likely to be better at preserving the information between the hidden states and the model output (compared to Scaled Dot-product); (ii) ablation studies indicate that Additive attention can actively learn to explain the importance of its input hidden representations; (iii) when attention values are nearly the same, the rank order of attention values is not consistent with the rank order of the mutual information(iv) Using Gumbel-Softmax with a temperature lower than one, tends to produce a more skewed attention score distribution compared to softmax and hence is a better choice for explainable design; (v) some building blocks are better at preserving the correlation between the ordered list of mutual information and attention weights order (for e.g., the combination of BiLSTM encoder and Additive attention). Our findings indicate that attention mechanisms do have the potential to function as a shortcut to model explanations when they are carefully combined with other model elements.