Conifer

该数据集名为Conifer，旨在提升大型语言模型（LLM）的指令跟随能力，通过提供具有复杂约束的多层次指令来实现此目标。经过一系列使用GPT-4的精细处理，该数据集精心挑选了35,613条指令。此外，该数据集采用渐进式学习方法，强调从易到难的训练过程，以解决LLM在处理复杂指令时遇到的挑战。规模上，数据集包含了35,613条指令，其任务是跟随具有复杂约束的指令。

This dataset, named Conifer, aims to enhance the instruction-following capabilities of Large Language Models (LLMs) by providing multi-level instructions with complex constraints. Through a series of meticulous curation processes powered by GPT-4, this dataset has carefully selected 35,613 instruction samples. Additionally, this dataset adopts a progressive learning approach, emphasizing an easy-to-difficult training progression to address the challenges that LLMs face when handling complex instructions. In terms of scale, the dataset contains 35,613 instruction instances, each requiring the model to follow complex-constrained instructions.