Jea Kwon

Bilinear representation mitigates reversal curse and enables consistent model editing

Erase or Hide? Suppressing Spurious Unlearning Neurons for Robust Unlearning

Textual Supervision Enhances Geospatial Representations in Vision-Language Models

Training Dynamics of Parametric and In-Context Knowledge Utilization in Language Models

Brain-inspired $L_p$-Convolution benefits large kernels and aligns better with visual cortex