# Representation and Reference Selection in Training-Free Synthetic Image Attribution

arXiv: https://arxiv.org/abs/2607.12052 · HTML: https://mlsanity.com/paper/2607.12052/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 17 of its 23 references that are in the map.
- First posted on arXiv 13 Jul 2026.

## Builds on (most cited first, 17 of 17)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Improved Baselines with Visual Instruction Tuning](https://mlsanity.com/paper/2310.03744/) (2023)
- [BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation](https://mlsanity.com/paper/2201.12086/) (2022)
- [Leveraging Frequency Analysis for Deep Fake Image Recognition](https://mlsanity.com/paper/2003.08685/) (2020)
- [DE-FAKE: Detection and Attribution of Fake Images Generated by Text-to-Image Generation Models](https://mlsanity.com/paper/2210.06998/) (2022)
- [Deepfake Network Architecture Attribution](https://mlsanity.com/paper/2202.13843/) (2022)
- [RepMix: Representation Mixing for Robust Attribution of Synthesized Images](https://mlsanity.com/paper/2207.02063/) (2022)
- [Detection, Attribution and Localization of GAN Generated Images](https://mlsanity.com/paper/2007.10466/) (2020)
- [Source Generator Attribution via Inversion](https://mlsanity.com/paper/1905.02259/) (2019)
- [Learning to Disentangle GAN Fingerprint for Fake Image Attribution](https://mlsanity.com/paper/2106.08749/) (2021)
- [Are CLIP features all you need for Universal Synthetic Image Origin Attribution?](https://mlsanity.com/paper/2408.09153/) (2024)
- [Forensic Self-Descriptions Are All You Need for Zero-Shot Detection, Open-Set Source Attribution, and Clustering of AI-generated Images](https://mlsanity.com/paper/2503.21003/) (2025)
- [Model Synthesis for Zero-Shot Model Attribution](https://mlsanity.com/paper/2307.15977/) (2023)
- [Training-free Source Attribution of AI-generated Images via Resynthesis](https://mlsanity.com/paper/2510.24278/) (2025)
- [Attribution as Retrieval: Model-Agnostic AI-Generated Image Attribution](https://mlsanity.com/paper/2603.10583/) (2026)
- [Who Made This? Fake Detection and Source Attribution with Diffusion Features](https://mlsanity.com/paper/2510.27602/) (2025)

## Abstract (arXiv)

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.