Inducing Artificial Uncertainty in Language Models

Sophia Hager, Simon Zeng, Nicholas Andrews

In safety-critical applications, language models should be able to characterize their uncertainty with meaningful probabilities. Many uncertainty quantification approaches require…