Instruction Mining: Instruction Data Selection for Tuning Large Language Models
This addresses the problem of optimizing finetuning efficiency for LLM developers, though it is incremental as it builds on existing finetuning methods.
The paper tackles the lack of standardized guidelines for selecting high-quality instruction-following datasets to finetune large language models, proposing InstructMining to automatically select premium data and achieving state-of-the-art performance on benchmarks like LLM-as-a-judge and Huggingface OpenLLM leaderboard.
Large language models (LLMs) are initially pretrained for broad capabilities and then finetuned with instruction-following datasets to improve their performance in interacting with humans. Despite advances in finetuning, a standardized guideline for selecting high-quality datasets to optimize this process remains elusive. In this paper, we first propose InstructMining, an innovative method designed for automatically selecting premium instruction-following data for finetuning LLMs. Specifically, InstructMining utilizes natural language indicators as a measure of data quality, applying them to evaluate unseen datasets. During experimentation, we discover that double descent phenomenon exists in large language model finetuning. Based on this observation, we further leverage BlendSearch to help find the best subset among the entire dataset (i.e., 2,532 out of 100,000). Experiment results show that InstructMining-7B achieves state-of-the-art performance on two of the most popular benchmarks: LLM-as-a-judge and Huggingface OpenLLM leaderboard.