We propose DiCo (Disentangled Concept Representation), a novel framework that achieves hierarchical and disentangled cross-modal alignment. DiCo introduces a shared slot-based representation, where each slot acts as a part-level anchor across modalities and is further decomposed into multiple concept blocks. This design enables the disentanglement of complementary attributes (e.g., color, texture, shape) while maintaining consistent part-level correspondence between image and text.

Giyeol Kim is currently a Ph.D. student in the Perceptual AI Lab at Chung-Ang University, advised by Prof. Chanho Eom. His research interests include agentic AI, multimodal large language models, and video surveillance and understanding systems, with a particular focus on building intelligent learning systems capable of capturing complex relationships across heterogeneous modalities. His recent research spans real-world video retrieval and anomaly detection, person search, and text-to-image person re-identification.