CT

Curt Tigges

Things Curt Says on Podcasts

Where to Find Them

Curt Tigges has been a guest on The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis .

Recently: “The Neuron Basis: Goodfire's Curt Tigges on Agentic Interp, Monosemanticity & Reverse Engineering” on The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis (November 2026).

What They Said

“Let's say you have a neuron that fires on clothing, it fires on exercise equipment, and it fires on dishes and on tablecloths. That seems very polysemantic, right? But what if the concept that the model has is things you might find at a garage sale? Then suddenly those things are unified, but you might never see it from the outside.” — Curt Tigges, The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis

Asked how to decide whether a neuron represents one concept or several, Curt Tigges argues the answer may depend on the model's own groupings, which humans can't see. He says the better question is whether you can describe everything a neuron represents.

The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis · 2026-11-10 Permalink → Listen →
The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis Around 56:16 into the episode
Curt Tigges

…in the data set. And then other metrics that we might use are: we say, given this sequence, can you tell me whether this neuron should activate in this sequence at all? Is it relevant at all? And at this one, we often see rates in the 90s because it's an easier task than token by token prediction.

Erik Torenberg

…around the nouns versus verbs thing. You know, maybe give me a little more of like a qualitative intuition for like, what are these, what are these paragraph length discussions? Somebody, I'm sure you've read, you know, thousands of them. What is the nature of these paragraph length descriptions?

Curt Tigges

They can vary widely between neurons. We see neurons doing so many different things. As you said, sometimes they'll contain multiple ideas, like this neuron will activate, you know, on oral health and marine environments, usually on the point at which someone decides to get oral surgery or at the point at which a specific animal in the marine environment was described. The description will read something like this. Most neurons do have something of a central concept that unifies them. But as I've looked at, you know, thousands of neuron descriptions and what they actually tend to represent, I think I've come to a somewhat different philosophical perception of things than simple monosemanticity or polysemanticity. So like monosemanticity or just like semanticity in general is this idea that a neuron or something else like a feature represents a single thing, a single concept. It's just detecting whether that's there. But in practice, I think it ends up much more complex. So each neuron can connect to thousands of other neurons upstream. And that provides a lot of information to this neuron. One thing that you might…

Erik Torenberg

So as you've looked at these circuits, do you have kind of a simple story around like, I mean, this sounds silly to say, but like kind of how it works? Like, I mean, what sort of motifs or, you know, moves stand out as sort of the notable primitives that could give us some intuition. I mean, I…

Curt Tigges

Well, some great questions in there.

Erik Torenberg

Only the big ones for you.

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.
“A neuron is not a fact. The fact is in the connections between neurons. And you can have exponentially more connections between neurons than you have individual neurons. So a set of 10 neurons networked together can represent way more than 10 facts, just because the number of possible paths between them is so high.” — Curt Tigges, The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis

The host asked where facts live if each neuron has roughly one core role, since a model with millions of neurons clearly stores far more facts than that. Goodfire's Curt Tigges answers that knowledge lives in the wiring, not in single neurons.

The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis · 2026-11-10 Permalink → Listen →
The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis Around 1:15:02 into the episode
Nathan Labenz

Let's maybe do a little lightning round if such a thing is possible in this context on just like how this work interacts with, or maybe should cause me to update my understanding of some other parts of what I

Erik Torenberg

…of always knew that was somewhat of a general, you know, an oversimplification. How would you update my thinking about facts? Like, do they, do we still think they live in the MLPs? Do they like this sort of one core role per neuron does leave me like, okay, then where are the facts in this thing?

Curt Tigges

Yeah, that's a frequent question because, you know, a model might have 5 million neurons, but no, way more than 5 million facts, right? But that's because you don't have a fact in a neuron. A neuron is not a fact, but the fact is in the connections between neurons. And you can have exponentially more connections between neurons than you have individual neurons. So, you know, a set of 10 neurons networked together can represent way more than 10 facts just because the number of possible paths between them is just so high. So yeah, from what we've seen, certainly we haven't found every single fact that's in the model and can say 100% are only in the MLPs, but we find so many facts in the MLPs just implemented by these connections that it seems like the natural place for them to be. Like the post-impressionism thing, you know, that's a whole bunch of artists and a whole bunch of concepts and in a smaller set of neurons than there are things that could be elaborated from that community.

Erik Torenberg

Do we see a different profile of factual recall degradation with sparsity as compared to like arc style tests? They can even do better with sparsity, reasoning they do worse. My gut would say factual would also be like significantly affected by sparsity, but what did you actually observe?

Curt Tigges

So we did test some MMLU data sets and we're able to recover performance on those pretty well. And that does test factual recall. We do have some tests that we're still working on for like medicine knowledge or chemistry knowledge. But for the most part, we really haven't seen much degradation.…

Erik Torenberg

Yeah, interesting. I guess that really just speaks to like localization, which then brings to mind the nature of experts. The old story about experts was the experts themselves are also not interpretable bad news. You know, it's, we can't really identify what role the experts are playing. It seems…

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.

Collections They Appear In

AI Research & Frontier Labs