Andy Phu


Preface


I research

1.

How LLMs can better understand Humans (context efficiency, user modelling)

2.

How Human-LLM interactions will look post-AGI (model welfare, alignment)

3.

How Humans can better understand LLMs (interp, safe... See more


Publications


SERUM: State Extraction and Refinement for User Modeling

Andy J. Phu, Karin de Langis, James Mooney, Khanh Chi Le, Dongyeop Kang. Accepted COLM 2026.


Blog