PhD Position in Post-Training of Diffusion Language Models (Prof. I. Bogunovic)

The Rhine AI group of Prof. Ilija Bogunovic in the Department of Mathematics and Computer Science offers a PhD position in the SNSF-funded project "Foundations of Post-Training for Reasoning in Diffusion Large Language Models". The doctoral researcher will develop reinforcement learning methods for post-training large diffusion-based language models, an emerging alternative to autoregressive LLMs, explore reasoning-aware generation strategies such as adaptive unmasking and credit assignment, and help build benchmarks for reasoning across domains including mathematical problem solving.

The group publishes at venues such as ICML, NeurIPS and ICLR. The 100% position starts in January 2027 or later; a Swiss PhD lasts four years. Candidates are encouraged to apply before 20 November 2026, but there is no fixed deadline.

Master's degree in computer science, mathematics, physics or a related field; exceptional academic record and solid mathematical foundation; strong coding skills and ability to run large-scale experiments; experience with LLMs, reinforcement learning or generative modelling is a plus; fluent English.