Joon Hyeok Kim

I am a graduate research assistant at GMLR Lab led by Prof. Jiatao Gu at the University of Pennsylvania.

My research journey is driven by a fundamental curiosity: “How do complex neural systems actually learn and generalize?” My current work focuses on the Mechanistic Interpretability of Deep Generative Models, specifically investigating how diffusion models learn algorithmic structures (e.g., modular addition and grokking).

Prior to joining UPenn, I spent three years as a Software Engineer at LG CNS, where I supported LG Electronics’ Global ERP (GERP) system. My background in Macroeconomics, combined with this engineering experience, has shaped a unique long-term goal: to build transparent and reliable World Models that can provide rigorous insights for complex decision-making systems, such as macroeconomic policy. I aim to bridge the gap between the black-box nature of deep learning and the high-stakes demand for accountability in critical real-world applications.