Adversarial Training (and Testing) | Stanford CS224U Natural Language Understanding | Spring 2021 скачать в хорошем качестве

Adversarial Training (and Testing) | Stanford CS224U Natural Language Understanding | Spring 2021

For more information about Stanford’s Artificial Intelligence professional and graduate programs, visit: https://stanford.io/ai To learn more about this course visit: https://online.stanford.edu/courses/c... To follow along with the course schedule and syllabus visit: https://web.stanford.edu/class/cs224u/ Professor Christopher Potts Professor and Chair, Department of Linguistics Professor, by courtesy, Department of Computer Science 0:00 Introduction 0:12 Overview 0:34 SWAG: Situations With Adversarial Generations 0:56 SWAG examples 1:47 Adversarial filtering for SWAG 2:57 Model accuracies under adversarial filtering 3:42 SWAG in the original BERT paper 4:26 HellaSWAG 5:26 Adversarial NLI: Dataset creation 6:55 Adversarial NL: Example Premise Hypothesis 7:24 Adversarial NL results 8:00 A vision for future development 8:59 Dynabench 9:39 Can adversarial training improve systems?