Aug, 2023

上下文对齐:在微调之前与简单语言模型聊天

TL;DRVanilla pretrained language model achieves improved alignment through in-context learning without changing model weights, resulting in comparable performance to strong baselines with alignment fine-tuning.