1 paper
Yubin Kim, Hyewon Jeong, Shan Chen +24
Hallucinations in foundation models arise from autoregressive training objectives that prioritize token-likelihood optimization over epistemic accuracy, fostering overconfidence an…