Surfacing Biases in Large Language Models using Contrastive Input Decoding

Open in new window