Lyndon Lam

I am currently a postbaccalaureate researcher in ML Foundations at the Kempner Institute. Previously, I graduated from Cal Poly Pomona.

Email  /  CV  /  Github

This is me.

News

• July 2024: Joined Kempner Institute.
• June 2023: Our paper was accepted in IROS 2023.

Research

I am interested in multimodal datasets and models. I am particularly interested in cross-modal transfer learning, where training a model on one modality (e.g., text-only QA) can improve performance on another modality (e.g., visual QA). Outside of multimodal learning but in a similar vein, I am also interested in cross-lingual transfer learning, such as training multilingual LLMs on English-only mathematical reasoning data and observing gains on Chinese math reasoning benchmarks. Broadly, I like training and advancing multimodal models, using methods that are efficient in both training and data.

Proposed a two-part model consisting of a classifier and an optimizer to generalize manipulation skills to novel object categories using verbs.

Template.