tinlab studies humans and machines, centered around topics like (linguistic) generalization, meaning, evaluation methods and design, and the nature of computation and representation that underlies language and cognition. tinlab members come from a wide range of academic backgrounds including Linguistics, Cognitive Science, Computer Science, Data Science, and Psychology.
tinlab is led by Najoung Kim, and is affiliated with the Department of Linguistics, Computer Science, and Computing & Data Science at Boston University. The lab is also a part of the SALAMI Consortium, a growing group of researchers at BU interested in computation and language.
Welcome new lab members: Yan, David, Max, and Jan!
The schedule for this semester’s invited speaker series is up!
“LLMs Learn Better In-Context from Rules than from Examples” has been accepted to EMNLP 2026! (with Xiang Fu, Seung-Min Cho, and Yukyung Lee)
“Does Episodic Memory Help Close the Lexical Frequency Gap in Sensitivity to Syntactic Contrasts? A Test Using Retrieval-Augmented Language Models” has been accepted to COLM 2026! Work with Jing Liu.
Preprint for “No Escape from Behavior in Evaluating Compositionality” is out on PsyArxiv.
Yulu has been awarded the Hariri Graduate Student Fellowship!
Najoung gave a talk titled Language models for language sciences at Stanford NLP. Slides here.
Do Language Models Track Entities Across State Changes? has been accepted to ICML 2026! Work with Peter Tang, Qiao Zhao, Gabriel Franco, Derry Wijaya, Aaron Mueller, and Sebastian Schuster.
tinlab @ BU