A arxiv.org Dola Decoding by Contrasting Layers Improves Factuality in Large Language Models 58 points johnsutor 2 years ago 43 comments
