Twitxy Saved

Tom McCoy

@RTomMcCoy

Assistant professor @YaleLinguistics. Studying computational linguistics, cognitive science, and AI. He/him.
New Haven, CT rtmccoy.com Joined December 2018
641 Following 5.1K Followers 2.2K Posts

Latest posts 5

Pinned post

Tom McCoy @RTomMcCoy
🤖🧠NEW PAPER🧠🤖
(The result of an 8-year project!)

LLMs seem very different from symbolic systems. Yet LLMs excel in symbolic domains (e.g., language/code/math). How do they do it?

Our finding: LLM representations have implicit symbolic structure!

Link in thread ⬇️
1/n
45 replies 270 reposts 1.4K likes 213K views View on X
Tom McCoy @RTomMcCoy
What makes LLMs so powerful? One factor might be that they can combine discrete and continuous variables, handling both the compositional side of the world and the statistical side of the world. Herbert's work gives an exciting illustration of this - check it out!

1/2
Herbert Zhou @zh_herbert_zhou Oct 24
What survives when we compress context into a steering vector ?

At #EMNLP2026, I'll present a way to causally edit continuous variables inside these vectors ⚙️, with a case study in linguistic knowledge.

Oral presentation @ Oct 25 11:00am F8 Interp — starting in 12h✨ A preview
View quoted post on X
Tom McCoy @RTomMcCoy
Najoung gave a fantastic talk. She's one of the people thinking most deeply about how the study of LLMs can inform human CogSci - her work proposes systematic principles where most of the discussion in the field is pretty anecdotal. Would highly recommend as a colloquium speaker!
Najoung Kim 🫠 @najoungkim Oct 22
had a great time giving a talk at the Yale Lingusitics colloquium about "beyond associative insights from AI" (thanks @RTomMcCoy for hosting), and had lots of fun conversations with the folks there! again featuring my favorite bridge figure:
View quoted post on X
1 replies 4 reposts 50 likes 2.2K views View on X