DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Models

Open in new window