Learning Using Generated Privileged Information by Text-to-Image Diffusion Models

TL;DR

This paper introduces LUGPI, a framework that uses text-to-image diffusion models to generate synthetic privileged information, enhancing text classification models without extra inference costs.

Contribution

It proposes a novel method to generate privileged information via diffusion models, improving text classification through multimodal teacher-student distillation.

Findings

Significant performance improvements on four datasets.

Effective use of synthetic images as privileged information.

No additional inference cost during deployment.

Abstract

Learning Using Privileged Information is a particular type of knowledge distillation where the teacher model benefits from an additional data representation during training, called privileged information, improving the student model, which does not see the extra representation. However, privileged information is rarely available in practice. To this end, we propose a text classification framework that harnesses text-to-image diffusion models to generate artificial privileged information. The generated images and the original text samples are further used to train multimodal teacher models based on state-of-the-art transformer-based architectures. Finally, the knowledge from multimodal teachers is distilled into a text-based (unimodal) student. Hence, by employing a generative model to produce synthetic data as privileged information, we guide the training of the student model. Our framework, called Learning Using Generated Privileged Information (LUGPI), yields noticeable performance gains on four text classification data sets, demonstrating its potential in text classification without any additional cost during inference.