SURF Mentoring

Potential projects/topics: This project focuses on helping undergraduate students develop a foundational understanding of large language models (LLMs). Students will explore how LLMs are trained at a high level, what they can and cannot do, and why evaluation of their behavior is an important research area. Potential topics include examining model outputs and comparing how different prompting strategies affect model responses.

Students may work on small-scale experiments such as analyzing model responses to structured tasks, designing simple evaluation benchmarks, or assessing how well model outputs align with human expectations. Example activities include qualitative analysis of model-generated text, basic statistical comparisons of outputs, and comparing strengths and limitations of LLMs across different use cases. By the end of the program, students will gain practical exposure to current AI research methods and develop an informed perspective on the capabilities and limitations of modern language models.

Potential skills gained: Understanding and evaluating large language models, experimental design, analyzing model outputs, critical thinking about AI limitations and ethics, and communicating research findings clearly.

Required qualifications or skills: This project is intended for Computer Science/AI majors. Students should be proficient in Python programming, familiar with machine learning concepts, and experienced with frameworks such as PyTorch and Hugging Face. Prior research experience is helpful but not required.

Direct mentor: Faculty/P.I., Post-doctorate, Graduate Student