Instruction tuning fine-tunes pre-trained Multi-modal Large Language Models (MLLMs) to handle real-world tasks. However, the rapid expansion of visual instruction datasets introduces data redundancy, leading to excessive computational costs. We propose a collaborative framework, DataTailor, which leverages three key principles--informativeness, uniqueness, and representativeness--for effective data selection. We argue that a valuable sample should be informative of the task, non-redundant, and represent the sample distribution (i.e., not an outlier). We further propose practical ways to score against each principle, which automatically adapts to a given dataset without tedious hyperparameter tuning. Comprehensive experiments on various benchmarks demonstrate that DataTailor achieves 101.3% of the performance of full-data fine-tuning with only 15% of the data, significantly reducing computational costs while maintaining superior results. This exemplifies the "Less is More" philosophy in MLLM development. The code and data is available in this \href{https://github.com/Yuqifan1117/DataTailor}{URL}.
@article{arxiv.2412.06293,
title = {Mastering Collaborative Multi-modal Data Selection: A Focus on Informativeness, Uniqueness, and Representativeness},
author = {Qifan Yu and Zhebei Shen and Zhongqi Yue and Yang Wu and Bosheng Qin and Wenqiao Zhang and Yunfei Li and Juncheng Li and Siliang Tang and Yueting Zhuang},
journal= {arXiv preprint arXiv:2412.06293},
year = {2025}
}