A Multiscale Visualization of Attention in the Transformer Model
June 12, 2019 ยท Declared Dead ยท ๐ Annual Meeting of the Association for Computational Linguistics
"No code URL or promise found in abstract"
Evidence collected by the PWNC Scanner
Authors
Jesse Vig
arXiv ID
1906.05714
Category
cs.HC: Human-Computer Interaction
Cross-listed
cs.CL,
cs.LG
Citations
673
Venue
Annual Meeting of the Association for Computational Linguistics
Last Checked
1 month ago
Abstract
The Transformer is a sequence model that forgoes traditional recurrent architectures in favor of a fully attention-based approach. Besides improving performance, an advantage of using attention is that it can also help to interpret a model by showing how the model assigns weight to different input elements. However, the multi-layer, multi-head attention mechanism in the Transformer model can be difficult to decipher. To make the model more accessible, we introduce an open-source tool that visualizes attention at multiple scales, each of which provides a unique perspective on the attention mechanism. We demonstrate the tool on BERT and OpenAI GPT-2 and present three example use cases: detecting model bias, locating relevant attention heads, and linking neurons to model behavior.
Community Contributions
Found the code? Know the venue? Think something is wrong? Let us know!
๐ Similar Papers
In the same crypt โ Human-Computer Interaction
R.I.P.
๐ป
Ghosted
R.I.P.
๐ป
Ghosted
Improving fairness in machine learning systems: What do industry practitioners need?
R.I.P.
๐ป
Ghosted
Identifying Stable Patterns over Time for Emotion Recognition from EEG
R.I.P.
๐ป
Ghosted
Questioning the AI: Informing Design Practices for Explainable AI User Experiences
R.I.P.
๐ป
Ghosted
Deep Learning for Sensor-based Human Activity Recognition: Overview, Challenges and Opportunities
R.I.P.
๐ป
Ghosted
Educational data mining and learning analytics: An updated survey
Died the same way โ ๐ป Ghosted
R.I.P.
๐ป
Ghosted
Language Models are Few-Shot Learners
R.I.P.
๐ป
Ghosted
PyTorch: An Imperative Style, High-Performance Deep Learning Library
R.I.P.
๐ป
Ghosted
XGBoost: A Scalable Tree Boosting System
R.I.P.
๐ป
Ghosted