Sheridan Feucht (they/them) -
feucht.s[at]northeastern.eduHello! I'm a third-year PhD student advised by David Bau and Byron Wallace at Northeastern University. Before that, I completed my B.S. in Computer Science at Brown University, where I worked with Ellie Pavlick in the LUNAR Lab. I'm originally from Calgary, Canada.
I do mechanistic interpretability with the goal of understanding how language models represent concepts. How do LLMs represent the deeper, meaningful parts of language? What kinds of conceptual representations do they learn? How can this inform our basic assumptions of how language works in general? See a list of my publications here.Recent updates
Jan 2026: I'll be visiting Goodfire as a Research Fellow from January through March! Excited to live in San Francisco for a bit.
Dec 2025: Check out my new blog post: Is AI Writing Still Nonsense?
Dec 2025: Gave an invited lightning talk at the NeurIPS Mech Interp Workshop [Equifinality: There's Often Multiple Explanations].
Nov 2025: New preprint by student Kerem Şahin on ICL without copying.
Oct 2025: Had a great time visiting Brown and giving an ANCOR talk on recent work [LLMs represent words, not just tokens].
Sep 2025: Gave an invited Zoom talk on recent work for Prof. Kyle Mahowald's seminar at UT Austin [LLMs represent words, not just tokens].
Sep 2025: Posters on concept arithmetic, as well as in-progress diffusion model work, were accepted to the NeurIPS Mech Interp Workshop.
July 2025: Our work, The Dual-Route Model of Induction, was accepted to COLM.
April 2025: Wrote my first blog post on syllogisms and intelligence.
I'm also a jazz drummer, and I do a bit of oil painting and creative writing. If you ever want to do music or art stuff together, feel free to reach out.
What's in my headphones: King Harvest (Has Surely Come) - The Band