3 papers
cs.LG2026
Geometry-Lite: Interpretable Safety Probing via Layer-Wise Margin Geometry
Woo Seob Sim, Yu Rang Park
Prompt-level safety probes for large language models use hidden-state representations to separate safe from unsafe prompts, but strong average detection performance does not explai…
cs.CV2026
Learning Relative Representations for Fine-Grained Multimodal Alignment with Limited Data
Shiwon Kim, Yu Rang Park
Multimodal pre-training demonstrates strong generalization performance, but this paradigm is often impractical in domains where paired data are scarce. A promising alternative is p…
cs.CV2026
MoECLIP: Patch-Specialized Experts for Zero-shot Anomaly Detection
Jun Yeong Park, JunYoung Seo, Minji Kang +1
The CLIP model's outstanding generalization has driven recent success in Zero-Shot Anomaly Detection (ZSAD) for detecting anomalies in unseen categories. The core challenge in ZSAD…