Laerdon Kim - Undergraduate Researcher at Cornell University
Hi! I'm Laerdon Kim, a second-year at Cornell University studying computer science. I'm interested in representation learning and encoders, conversational modeling and simulation, and proactive + personalized human–AI interaction. My research advisors are Professor Cristian Danescu-Niculescu-Mizil and Professor Lillian Lee.
research interests
I am curious about, broadly, reimagining interaction paradigms around AI systems, which I hope to investigate in two ways:
- I am interested in continual learning from human feedback as a practical path toward systems that become more personalized and aligned. I care about proactive applications that help users think, reflect, and act more effectively over time. This points toward interfaces and systems that are more context-aware, forward-looking, and supportive of sustained human goals.
- I want to build models that can reliably understand, simulate, and steer conversational dynamics. This includes predicting how interactions unfold, modeling the structure of dialogue, and designing systems that can intervene constructively in real time. I see education as a particularly compelling application area, and I am inspired by work on social simulation and agent-based interaction.
I am also interested in how concepts are represented in models, and in the limits of the autoregressive paradigm as a foundation for learning and reasoning. I am especially drawn to work on multimodal representation alignment, the Platonic Representation Hypothesis, and information-theoretic perspectives on language modeling.
what i'm up to
I recently co-authored an EMNLP 2025 paper on lexical semantic change! Read more here.
for fun
- I produce music in Ableton, occasionally contributing samples of my cello.
- I sketch. I have a lot of MUJI pens, too.
- I like to grill, and bake cupcakes, pies, and custard-related desserts. I once made an algae-colored pumpkin pie.
- In summer 2024, I started a blog on Substack.