EMNLP Findings2020
Inserting Information Bottleneck for Attribution in Transformers
Abstract
Pretrained transformers achieve the state of the art across tasks in natural language processing, motivating researchers to investigate their inner mechanisms. One common direction is to understand what features are important for prediction. In this paper, we apply information bottlenecks to analyze the attribution of each feature for prediction on a black-box model. We use BERT as the example and evaluate our approach both quantitatively and qualitatively. We show the effectiveness of our method in terms of attribution and the ability to provide insight into how information flows through layers. We demonstrate that our technique outperforms two competitive methods in degradation tests on four datasets. Code is available at https://github.com/bazingagin/IBA.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
2 top-tier papers have cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
Cited by top-tier papers2
- Visual Explanations of Image-Text Representations via Multi-Modal Information Bottleneck AttributionYing Wang, Tim G. J. Rudner, Andrew Gordon WilsonNeurIPS 2023 · 51 citations
- IF-Prune: Information-Flow Guided Token Pruning for Efficient Vision-Language ModelsGuohao Sun, Yufei Wang, Sizhuo Ma, Yuege Xie et al.CVPR 2026
Builds on2
Related papers
- Fine-Grained Neural Network Explanation by Identifying Input Features with Predictive InformationYang Zhang, Ashkan Khakzar, Yawei Li, Azade Farshad et al.NeurIPS 2021 · 33 citations
- Self-Attention Attribution: Interpreting Information Interactions Inside TransformerYaru Hao, Li Dong, Furu Wei, Ke XuAAAI 2021 · 282 citations
- Comprehensive Information Bottleneck for Unveiling Universal Attribution to Interpret Vision TransformersJung-Ho Hong, Ho-Joong Kim, Kyu-Sung Jeon, Seong-Whan LeeCVPR 2025
- Influence Patterns for Explaining Information Flow in BERTKaiji Lu, Zifan Wang, Piotr Mardziel, Anupam DattaNeurIPS 2021 · 22 citations
- Measuring the Mixing of Contextual Information in the TransformerJavier Ferrando, Gerard I. Gállego, Marta R. Costa-jussàEMNLP 2022 · 18 citations
