Arman Cohan Research Group

Yale University

How can language models and representation learning be adapted to long-context, multi-document, and specialized-domain tasks while improving reliability and evaluation? The Yale NLP Lab studies language modeling, representation learning, retrieval, and domain-specific applications across mechanistic interpretability and evaluation. Researchers investigate capabilities and reasoning of large language models, retrieval-augmented methods, and knowledge-intensive environments using experiments and modeling. The group develops pretraining and summarization techniques for long and multi-document inputs and evaluates few-shot and faithfulness metrics for real-world NLP tasks. Lab members publish in major NLP and ML venues and maintain an active research and student training program.