Evaluating External Knowledge
Daniel and Sewon discuss the evaluation of external knowledge in machine learning models and how it can be validated without retraining the model. They explore the use of Wikipedia corpus as an example and its potential for updating knowledge without the need for retraining. Title: Verifying Claims and Lexical Bias Topics: Claims verification, Lexical bias Summary: Sewon explains the process of creating a data set that contains natural and unbiased claims. Daniel raises questions about the carefulness of benchmark verification and Sewon unpacks the process of creating realistic and challenging false claims through crossover interpretations. They highlight the importance of complete understanding of evidence for claim verification.In this clip
From this podcast
The Gradient
Sewon Min: The Science of Natural Language
Related Questions