Our Paper (MoECLIP) Accepted at CVPR 2026

We are pleased to announce that our paper, "MoECLIP", has been accepted at CVPR 2026.

MoECLIP proposes a Mixture-of-Experts architecture for Zero-Shot Anomaly Detection, achieving patch-level adaptation by dynamically routing each image patch to a specialized LoRA expert.

Jun Yeong Park, JunYoung Seo, Minji Kang, Yu Rang Park. MoECLIP: Patch-Specialized Experts for Zero-shot Anomaly Detection. CVPR2026 (accepted)

🔗 arXiv |