One-Shot Learning as Instruction Data Prospector for Large Language Models

TL;DR

This paper introduces extsc{Nuggets}, a method that uses one-shot learning to select high-quality instruction data, improving large language model tuning by filtering out noise and enhancing performance.

Contribution

extsc{Nuggets} provides an efficient way to identify valuable instruction examples for tuning large language models, outperforming traditional data scaling methods.

Findings

Selective data improves model performance significantly.

Top 1 ext% of curated data outperforms full datasets.

Method effective across multiple benchmarks.

Abstract

Contemporary practices in instruction tuning often hinge on enlarging data scaling without a clear strategy for ensuring data quality, inadvertently introducing noise that may compromise model performance. To address this challenge, we introduce \textsc{Nuggets}, a novel and efficient methodology that leverages one-shot learning to discern and select high-quality instruction data from extensive datasets. \textsc{Nuggets} assesses the potential of individual instruction examples to act as effective one-shot learning instances, thereby identifying those that can significantly improve performance across diverse tasks. \textsc{Nuggets} utilizes a scoring system based on the impact of candidate examples on the perplexity of a diverse anchor set, facilitating the selection of the most advantageous data for instruction tuning. Through comprehensive evaluations on two benchmarks, including MT-Bench and Alpaca-Eval, we show that instruction tuning with the top 1\% of examples curated by \textsc{Nuggets} substantially outperforms conventional methods employing the entire dataset.