Pre-trained Language Models Return Distinguishable Probability Distributions to Unfaithfully Hallucinated Texts

HILM
Re-assign community