Tsung-Huan Yang

I'm an incoming MSCS student at Georgia Tech. My research centers on Trustworthy AI, with the goal of building safe, interpretable, and reliable AI systems. I'm currently exploring topics such as mechanistic interpretability, automated/multimodal red-teaming, and deceptive alignment. I join Georgia Tech after a two-year RAship at Academia Sinica, where I worked with Dr. Lun-Wei Ku on building culture-aware safety classifiers and generating adversarial attacks for LLM safety/security assessment. Before that, I've worked on speech processing under Prof. Hung-yi Lee and model compression with Prof. Hao Tang.