How Do We Know a Clinical Summary is Good? Having spent years working on summarization evaluation in my PhD (before LLM era), I know just how hard it is: There is no single “correct” summary—different summaries may emphasize different relevant points, and human judgment varies.

We’re excited to announce that our paper Anchored Answers: Unravelling Positional Bias in GPT-2’s Multiple-Choice Questions on has been accepted to an upcoming ACL 2025! In this study, we investigate how GPT-2 internally represents and reasons over Multiple-Choice Questions (MCQA) tasks.

General Instructions for Giving a Job Talk at LARK Lab At LARK Lab, we value engaging discussions, deep technical insights, and collaborative learning. A job talk is an opportunity for us to learn about your research, thought process, and how you approach problems.

We are excited to announce the launch of our Reinforcement Learning (RL) Seminar, a biweekly reading and discussion series focused on foundational RL concepts and applications. With several ongoing RL-related projects in our lab and a growing need for structured training among our trainees, we are introducing this seminar-style reading club to foster a deeper understanding of RL principles.

Dr. Yanjun Gao recently spoke at a campus conference on “Engaging with AI”, where she shared insights on the evolving role of AI in research, healthcare, and education. The talk explored how AI innovations, particularly large language models (LLMs), are transforming scientific discovery and decision-making.