Our research focuses on understanding behaviors of large language models (LLMs), adapting them effectively to knowledge-intensive reasoning tasks, and aligning them safely with users from diverse backgrounds.
Research themes
- Multilingual NLP and Cultures: Democratizing inclusive LLMs across languages and cultures while enhancing pluralistic alignment.
- Efficient Multimodal LLMs: Developing efficient multimodal LLMs for high-stakes applications in education and healthcare.
- Cognitive Language Agents: Unifying cognitive concepts and languages to enable complex reasoning, planning and human behavior simulation.
- Evaluation and Mechanistic Interpretation: Exploring robust methods for evaluating and interpreting black-box foundation models.
Lab Outing - Summer 2026
Lab Outing - Summer 2026
Lab Outing - Spring 2024
Lab Outing - Spring 2023
Sponsors
We are grateful to the following organizations for supporting our research:
- National Science Foundation (CISE/IIS core)
- National Institutes of Health (NLM R01, NIBIB R01)
- Department of Defense
- IARPA (Video LINCS)
- UW ICTR
- Wisconsin Alumni Research Foundation
- American Family Insurance
- Coefficient Giving (previously Open Philanthropy)
- Microsoft
- NVIDIA