DataInf: Efficiently Estimating Data Influence in LoRA-tuned LLMs and Diffusion Models

TL;DR

DataInf is a novel, efficient influence approximation method designed for large-scale generative models, significantly reducing computational costs while accurately identifying influential training data points and mislabeled data.

Contribution

It introduces a closed-form influence approximation tailored for large models and parameter-efficient fine-tuning, outperforming existing methods in speed and accuracy.

Findings

DataInf accurately approximates influence scores in large models.

It is orders of magnitude faster than existing influence computation algorithms.

DataInf effectively identifies influential and mislabeled data points in various models.

Abstract

Quantifying the impact of training data points is crucial for understanding the outputs of machine learning models and for improving the transparency of the AI pipeline. The influence function is a principled and popular data attribution method, but its computational cost often makes it challenging to use. This issue becomes more pronounced in the setting of large language models and text-to-image models. In this work, we propose DataInf, an efficient influence approximation method that is practical for large-scale generative AI models. Leveraging an easy-to-compute closed-form expression, DataInf outperforms existing influence computation algorithms in terms of computational and memory efficiency. Our theoretical analysis shows that DataInf is particularly well-suited for parameter-efficient fine-tuning techniques such as LoRA. Through systematic empirical evaluations, we show that DataInf accurately approximates influence scores and is orders of magnitude faster than existing methods. In applications to RoBERTa-large, Llama-2-13B-chat, and stable-diffusion-v1.5 models, DataInf effectively identifies the most influential fine-tuning examples better than other approximate influence scores. Moreover, it can help to identify which data points are mislabeled.