current Mechanistic interpretability of VLA models What do vision-language-action models represent, and how is their latent space organized? World models for robotics 3D geometry-aware representations for world models and world action models. Neural Collapse and representation geometry How last-layer features and classifier weights organize, and what breaks when idealized assumptions fail. Disentangled and object-centric representations Consistent symmetry representations, layer-wise factor separation, and slot attention under continual learning. past Imbalanced and continual learning Margin balancing, stable replay, and class-wise interference in the last layer. Text mining for news issue analysis Hierarchical and incremental clustering of news articles, Korean word similarity, and applied data mining.
