TAGCOS: Task-agnostic Gradient Clustered Coreset Selection for Instruction Tuning Data

TL;DR

TAGCOS is a novel method that efficiently selects a small, highly informative subset of instruction tuning data using gradient clustering, significantly reducing data volume while maintaining high model performance.

Contribution

It introduces a task-agnostic coreset selection method based on gradient clustering, addressing diversity and efficiency challenges in instruction dataset selection.

Findings

Selecting only 5% of data achieves comparable performance to full dataset.

TAGCOS outperforms other unsupervised data selection methods.

The method is effective across diverse instruction datasets.

Abstract

Instruction tuning has achieved unprecedented success in NLP, turning large language models into versatile chatbots. However, the increasing variety and volume of instruction datasets demand significant computational resources. To address this, it is essential to extract a small and highly informative subset (i.e., Coreset) that achieves comparable performance to the full dataset. Achieving this goal poses non-trivial challenges: 1) data selection requires accurate data representations that reflect the training samples' quality, 2) considering the diverse nature of instruction datasets, and 3) ensuring the efficiency of the coreset selection algorithm for large models. To address these challenges, we propose Task-Agnostic Gradient Clustered COreset Selection (TAGCOS). Specifically, we leverage sample gradients as the data representations, perform clustering to group similar data, and apply an efficient greedy algorithm for coreset selection. Experimental results show that our algorithm, selecting only 5% of the data, surpasses other unsupervised methods and achieves performance close to that of the full dataset.