We study the connections between language, interaction, and intelligence. Our lab develops methods for interactive machine learning, explainability and interpretability of LLMs, and impactful interdisciplinary applications of NLP.
Intelligence does not arise in isolation, it is an evolving, interactive phenomenon. People use language not just to communicate, but to think, to plan and collaborate. While we have been interested in the connection between machine learning and human language for long, LLMs have made this space fundamentally more interesting. For example, Chain-of-Thought prompting unlocks complex abilities in LLMs, allowing them to circumvent architectural limitations (e.g., computational budgets due to fixed Transformer depth) much like humans externalize cognitive burdens through writing, structured reasoning, and collaboration. We believe these analogies reveal something about the fundamental role of language in thought and learning.
Shashank to give invited talk on 'Adapting Narrative to Adapt Minds' at WNU 2024
01 Nov 2024Three papers accepted at EMNLP 2024!
01 Sep 2024Welcome Akshan, Jane, Rajeev and Wisdom!
01 Feb 2024Shashank's invited lecture on 'A brief history of AI' at Honors Carolina
06 Oct 2023Five papers accepted to EMNLP 2023! Check out the pre-prints on the Publications page!