Год публикации
Артем Шелманов Михаил Бурцев Кирилл Федянин Максим Панов Манвел Аветисян Леонид Жуков Александр Панченко Артем Важенцев Глеб Кузмин Аким Цвигун Евгений Цымбалов Глеб Гусев

Uncertainty Estimation of Transformer Predictions for Misclassification Detection


Uncertainty estimation (UE) of model predictions is a crucial step for a variety of tasks such as active learning, misclassification / adversarial attack / out-of-distribution detection, etc. Most of the works on modeling the uncertainty of deep neural networks evaluate these methods on image classification tasks. Little attention has been paid to UE in natural language processing. To fill this gap, we perform a vast empirical investigation of state-of-the-art UE methods for Transformer models on misclassification detection in named entity recognition and text classification tasks and propose two computationally efficient modifications, one of which improves the state of the art and outperforms computationally intensive methods.

Присоединяйтесь к AIRI в соцсетях