Unsupervised Synthetic Image Attribution: Alignment and Disentanglement

2026-01-30

AI Summary

This work addresses the critical challenge of accurately attributing synthetic images to their original training concepts in the absence of paired annotations—a key requirement for copyright protection and model transparency. The authors propose an unsupervised attribution method that leverages contrastive self-supervised learning to align concepts across domains and incorporates an Infomax loss to encourage disentangled representations. Theoretical analysis reveals that the proposed framework is equivalent to a decomposition of the objective in canonical correlation analysis. Evaluated on the real-world AbC benchmark, the method achieves state-of-the-art performance, significantly outperforming existing supervised attribution approaches under the unsupervised setting, thereby demonstrating its effectiveness and potential.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Unsupervised & Self-Supervised Learning
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

As the quality of synthetic images improves, identifying the underlying concepts of model-generated images is becoming increasingly crucial for copyright protection and ensuring model transparency. Existing methods achieve this attribution goal by training models using annotated pairs of synthetic images and their original training sources. However, obtaining such paired supervision is challenging, as it requires either well-designed synthetic concepts or precise annotations from millions of training sources. To eliminate the need for costly paired annotations, in this paper, we explore the possibility of unsupervised synthetic image attribution. We propose a simple yet effective unsupervised method called Alignment and Disentanglement. Specifically, we begin by performing basic concept alignment using contrastive self-supervised learning. Next, we enhance the model's attribution ability by promoting representation disentanglement with the Infomax loss. This approach is motivated by an interesting observation: contrastive self-supervised models, such as MoCo and DINO, inherently exhibit the ability to perform simple cross-domain alignment. By formulating this observation as a theoretical assumption on cross-covariance, we provide a theoretical explanation of how alignment and disentanglement can approximate the concept-matching process through a decomposition of the canonical correlation analysis objective. On the real-world benchmarks, AbC, we show that our unsupervised method surprisingly outperforms the supervised methods. As a starting point, we expect our intuitive insights and experimental findings to provide a fresh perspective on this challenging task.

Problem

• Unsupervised Attribution
• Synthetic Image
• Concept Alignment
• Copyright Protection
• Model Transparency

Innovation

• unsupervised attribution
• contrastive self-supervised learning
• representation disentanglement
• synthetic image alignment
• Infomax loss