Pascal Jr. Tikeng Notsawo

I am a third-year PhD candidate at Mila and Université de Montréal, supervised by Irina Rish, Guillaume Rabusseau, and Guillaume Dumas. My previous research focused on generalization, optimization dynamics, and representation learning in neural networks, including theoretical and empirical studies of grokking and implicit bias in deep learning. More recently, my research has shifted toward AI safety and alignment, with a focus on LLM unlearning, and I am currently working with Tara Research on studying and improving honesty in language models.

I earned a Master of Engineering in Computer Science from National Advanced School of Engineering, Yaoundé, Cameroon. Before starting my PhD, I was a Visiting Student Researcher at Université de Montréal and a Research Intern at Mila, where I worked with Yoshua Bengio and Dianbo Liu. I also contributed to language technology for African languages and conducted research on the detection and mitigation of discriminatory language in online text.