Jens Tuyls

Senior Researcher at Microsoft Research NYC

prof_pic.jpg

My research generally focuses on reinforcement learning and natural language processing, and more recently I’ve started thinking about self-improvement in AI agents.

Bio: I recently graduated from the Princeton NLP group, where I was advised by Professor Karthik Narasimhan. My research focused on building decision-making agents, mainly using ideas from reinforcement learning (RL), imitation learning (IL), and natural language processing (NLP). To this end, I also closely collaborated with Professor Benjamin Eysenbach as part of the Princeton RL Lab.

Before grad school, I studied Computer Science and Engineering at UC Irvine, where I was fortunate to work with Professor Sameer Singh on interpretability in NLP, and with Professor Stephan Mandt and Professor Mike Pritchard on AI for climate.

news

Jan, 2026 My MSR internship project Representation-Based Exploration for Language Models: From Test-Time to Post-Training was accepted to ICLR 2026!
Jan, 2026 Our new LLM post-training algorithm called RepExp was open-sourced into the verl-recipe repo.
Oct, 2025 Completed internship at Microsoft Research, NYC, where I worked with Dylan Foster, Akshay Krishnamurthy, and Jordan Ash! Check out our recent preprint on “Representation-Based Exploration for Language Models: From Test-Time to Post-Training”.
Oct, 2025 Our work “Can a MISL Fly? Analysis and Ingredients for Mutual Information Skill Learning” was accepted as Oral at ICLR 2025!
Oct, 2024 Our work “Can a MISL Fly? Analysis and Ingredients for Mutual Information Skill Learning” was accepted to the IMOL workshop at NeurIPS.