EUVIMEDEUVIMED entdecken
Crossref

CausalCLIP: Causally-Informed Feature Disentanglement and Filtering for Generalizable Detection of Generated Images

Bo Liu, Qiao Qin, Qinghui He

Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (9) · 7069-7077

DOI: 10.1609/aaai.v40i9.37642 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Bo Liu und 2 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The rapid advancement of generative models has increased the demand for generated image detectors capable of generalizing across diverse and evolving generation techniques. However, existing methods, including those leveraging pre-trained vision-language models, often produce highly entangled representations, mixing task-relevant forensic cues (causal features) with spurious or irrelevant patterns (non-causal features), thus limiting generalization. To address this issue, we propose CausalCLIP, a framework that explicitly disentangles causal from non-causal features and employs targeted filtering guided by causal inference principles to retain only the most transferable and discriminative forensic cues. By modeling the generation process with a structural causal model and enforcing statistical independence through Gumbel-Softmax-based feature masking and Hilbert-Schmidt Independence Criterion (HSIC) constraints, CausalCLIP isolates stable causal features robust to distribution shifts. When tested on unseen generative models from different series, CausalCLIP demonstrates strong generalization ability, achieving improvements of 6.83% in accuracy and 4.06% in average precision over state-of-the-art methods.

Zitieren

APA

Liu, B., Qin, Q., & He, Q. (2026). CausalCLIP: Causally-Informed Feature Disentanglement and Filtering for Generalizable Detection of Generated Images. Proceedings of the AAAI Conference on Artificial Intelligence, 40(9), 7069-7077. https://doi.org/10.1609/aaai.v40i9.37642

Vancouver

Liu B, Qin Q, He Q. CausalCLIP: Causally-Informed Feature Disentanglement and Filtering for Generalizable Detection of Generated Images. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(9):7069-7077. doi:10.1609/aaai.v40i9.37642

BibTeX
@article{Liu2026,
  author = {Liu, Bo and Qin, Qiao and He, Qinghui},
  title = {CausalCLIP: Causally-Informed Feature Disentanglement and Filtering for Generalizable Detection of Generated Images},
  journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
  year = {2026},
  volume = {40},
  number = {9},
  pages = {7069--7077},
  doi = {10.1609/aaai.v40i9.37642}
}
Kostenlos starten →