Diverse and Fine-Grained Instruction-Following Ability Exploration with Synthetic Data

Zihui Gu, Xingwu Sun, Fengzong Lian +3

Instruction-following is particularly crucial for large language models (LLMs) to support diverse user requests. While existing work has made progress in aligning LLMs with human p…