Join the lab

Students in the lab ask how language models learn, reason, and behave, and how we can understand them better. They develop new methods, investigate unexpected model behavior, and test their ideas with careful experiments. The work spans natural-language explanations, interpretability, interactive learning, and the effects of language on model behavior.

Some problems we find interesting.

• Explaining a model’s behavior when only its inputs and outputs are visible.
• Causal analysis of how LLMs reason, and of when they commit to a decision.
• Models that learn by asking questions.
• How language shapes behavior when models interact with each other and with people.

Strong fundamentals in machine learning, mathematics, and programming.

Curiosity about how language models work, and why, beyond how well.

Mutual interest in the lab’s research thrusts: interpretability and explainability of LLMs, faithful reasoning, learning from language and interaction, and language, cognition, and social behavior.

Please fill out our interest form. You do not need to email the PI separately.

Interest form ↗

Apply through the UNC Computer Science graduate admissions and mention our lab and your research interests in your application.

Undergraduates have led more than ten papers at major ML and NLP conferences, including orals at ACL 2025 and EMNLP 2026. We welcome motivated students with strong technical foundations who want to take on a substantial research project.