Brian McCallion

Independent researcher working on foundational models of large language model behaviour and alignment. My work focuses on mechanistic explanations of hallucination, coherence, and failure modes using systems theory, information theory, and learning dynamics. I am currently developing a boundary-mediated framework for understanding inference and learning in LLMs, with an emphasis on testable predictions and alignment-relevant design patterns.

Personal Website

-

LinkedIn

-

Twitter / X

-

LessWrong

-

EA Forum

-

Projects

No grants recorded.