Representation and Reference Selection in Training-Free Synthetic Image Attribution

2026-07-13

AI Summary

This study addresses the coupling between representation space and reference sample selection in training-agnostic synthetic image provenance. It systematically investigates the impact of intermediate-layer features from pretrained models (CLIP, DINOv2) and three reference selection strategies—arbitrary, semantic alignment, and resynthesis—on provenance performance. The findings reveal that intermediate-layer representations achieve an optimal trade-off between preserving generator-discriminative cues and maintaining semantic consistency. Semantic constraints substantially mitigate query–reference mismatch: resynthesis proves most effective under low reference budgets, while semantic alignment offers the best cost–performance balance at moderate budgets. Building on these insights, the work proposes budget-adaptive strategies for constructing optimal reference sets.

Abstract

Synthetic image attribution aims at identifying the generator responsible for a given AI-generated image. Training-free reference-based attribution methods are easily scalable, since newly emerging generators can be incorporated by adding source-specific references rather than retraining a task-specific classifier. Their performance depends on two coupled factors: the representation space used for comparison and the way source-specific references are constructed. However, the interaction between these two factors remains largely unexplored. In this paper, we provide a controlled analysis of this interaction using references and off-the-shelf pretrained representations. We study representations extracted from different layers of CLIP and DINOv2, along with three reference selection methods with varying semantic constraints: arbitrary, semantically aligned, and resynthesis-based references. Our results show that attribution accuracy consistently peaks at intermediate representation levels, indicating that source-discriminative cues are more accessible before strong semantic abstraction dominates. We further show that intermediate representations are not completely semantically neutral, making reference selection critical: semantically constrained references reduce query-reference mismatch and improve attribution, especially under limited reference budgets. Resynthesis is most useful in low-reference regimes, while semantically aligned references provide a better accuracy-cost trade-off when a moderate-sized reference pool is available. Our findings show that training-free reference-based attribution should be understood as the interaction between where images are compared, how the reference set is constructed, and how many references are available.

Problem

• synthetic image attribution
• representation space
• reference selection
• training-free
• source identification

Innovation

• training-free attribution
• intermediate representations
• reference selection
• synthetic image provenance
• semantic alignment