Expanding the Role of Diffusion Models for Robust Classifier Training

2026-02-23

AI Summary

This work proposes a novel approach to enhance the robustness of image classifiers in adversarial training by jointly leveraging synthetic data generated by diffusion models and their internal representations. In contrast to prior methods that utilize only synthetic samples, this study is the first to systematically incorporate intermediate features from diffusion models as auxiliary supervision signals. The authors demonstrate that these internal representations exhibit complementary properties to the synthetic data in feature learning, fostering more disentangled and discriminative representations. Extensive experiments on CIFAR-10, CIFAR-100, and ImageNet show consistent and significant improvements in model robustness, underscoring the unique value of diffusion model representations in adversarial training.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Adversarial Learning & Robustness
• Natural Language Processing: Safety and Robustness

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Incorporating diffusion-generated synthetic data into adversarial training (AT) has been shown to substantially improve the training of robust image classifiers. In this work, we extend the role of diffusion models beyond merely generating synthetic data, examining whether their internal representations, which encode meaningful features of the data, can provide additional benefits for robust classifier training. Through systematic experiments, we show that diffusion models offer representations that are both diverse and partially robust, and that explicitly incorporating diffusion representations as an auxiliary learning signal during AT consistently improves robustness across settings. Furthermore, our representation analysis indicates that incorporating diffusion models into AT encourages more disentangled features, while diffusion representations and diffusion-generated synthetic data play complementary roles in shaping representations. Experiments on CIFAR-10, CIFAR-100, and ImageNet validate these findings, demonstrating the effectiveness of jointly leveraging diffusion representations and synthetic data within AT.

Problem

• diffusion models
• robust classifier training
• adversarial training
• representation learning
• synthetic data

Innovation

• diffusion representations
• adversarial training
• robust classification
• synthetic data
• feature disentanglement