NILE: Internal Consistency Alignment in Large Language Models

Minda Hu, Qiyuan Zhang, Yufei Wang +7

As a crucial step to enhance LLMs alignment with human intentions, Instruction Fine-Tuning (IFT) has a high demand on dataset quality. However, existing IFT datasets often contain…