'of' in university names
A neuronOne of thousands of tiny dials inside each layer. It reads the text so far and produces a number, its activation, that nudges what the model writes next. in layerOne step in the stack the text passes through. Qwen3-0.6B has 28 layers, each with 3,072 neurons. 16 of Qwen3-0.6B. Its strongest snippets are very consistent: they mostly peak on the same token. It is featuredA neuron picked for the full treatment: steering outputs and a written explanation. About 3,000 of the 86,000 neurons are featured., so it has steering outputs.
What this neuron does
This neuron lights up on the word 'of' when it appears in a university name, such as 'University of Arizona' or 'University of Chicago'. Turning it up didn't produce anything about universities, so it does not clearly change what the model writes.
Written by Claude from this neuron's activations and steering outputs. Confidence: medium. Think it is wrong? Flag it (opens a prefilled issue on GitHub).
Fires most on
The text windows where this neuron's activationThe number a neuron produces for a token. Large positive means it fired hard; negative means it pushed the other way. was highest. Amber marks positive activation; the outlined tokenA piece of text the model reads at once: often a word, sometimes part of a word or a punctuation mark. is the peak. Hover a token for its strength.
-
University of Hamburg in Germany, in collaboration with colleagues from the University of Aarhus in Denmark, have synthesized a new superconducting material by growing a
-
's Kavli Institute for Particle Astrophysics and Cosmology, in collaboration with Lucio Mayer of the Swiss Federal Institute of Technology in Zurich and the University of Zurich, Chiara Mastropietro of the University of Munich in Germany,
-
. Thomas Pichler from the University of Vienna report a method for the bulk production of long acetylenic linear carbon
-
in 1972, a BS in Physics from the University of California, Irvine in 1978 and a PhD in Astronomy from the University of Arizona in 1982. His research interests have included the linear and
-
and seismologist at the University of Arizona. The researchers suspect that a previously discovered region of mantle downwelling a
-
33, where he continued his study of social science and mathematics. Herbert Simon earned both his B.A. and Ph.D. in political science from the University of Chicago.⏎- Full Name⏎- Herbert A Simon⏎- APA Award for D
SteeringAdding a push along a neuron while the model writes, to see how its output changes. Turning it up or down.
Push this neuron up or down while the model writes, and watch what changes. The outputs were generated ahead of time, so nothing runs in your browser.
Here is something I have been thinking about lately: I have been thinking about lately, and I have been thinking about lately, and I have been thinking about lately, and I have been thinking about lately, and I have been thinking about lately, and I have been thinking about lately, and
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
, and then watch the current year Address and see if their expectations were met. In this State of the Union...⏎7th - 12th Social Studies & History⏎President Obama's State of the Union Address 2011⏎
-
networks called Alice, Bob, and Eve.⏎Ref: Learning to Protect Communications with Adversarial Neural Cryptography. arXiv - Computer Science > Cryptography and Security (21 October 2016) | arXiv:
-
as in a discontinuity analysis on observations near the borders between anglophone and francophone countries.⏎- Colonial rule⏎ASJC Scopus subject areas⏎- Economics and Econometrics⏎- Organizational Behavior and Human Resource Management<|endoftext|>- About AST
-
between model and experiment was observed.⏎ASJC Scopus subject areas⏎- Electronic, Optical and Magnetic Materials⏎- Ceramics and Composites⏎- Condensed Matter Physics⏎- Materials Chemistry<|endoftext|>Humans are intimately connected with the physical environment, even
Connections
Show on mapWhich neuronsOne of thousands of tiny dials inside each layer. It reads the text so far and produces a number, its activation, that nudges what the model writes next. in earlier layersOne step in the stack the text passes through. Qwen3-0.6B has 28 layers, each with 3,072 neurons. feed this one, and which later ones it feeds. Amber links excite, blue links suppress; longer bars are stronger.
Fed by earlier layers
- L15:2688 excites
- L9:218 weakly excites
- L15:124 weakly suppresses
- L15:474 weakly excites
- L4:1059 weakly excites
- L14:1755 weakly suppresses
- L12:2843 weakly excites
- L15:1499 weakly excites
- L6:1926 weakly excites
- L14:1569 weakly suppresses
- L13:2269 weakly suppresses
- L15:130 weakly suppresses
- L8:2101 weakly excites
- L15:1504 weakly suppresses
- L15:488 weakly excites
- L10:1521 weakly suppresses
- L12:345 weakly suppresses
- L15:954 weakly excites
- L2:2135 weakly excites
- L0:1967 weakly suppresses
- L15:2041 weakly suppresses
- L5:612 weakly suppresses
- L15:2637 weakly excites
- L15:2279 weakly excites
Feeds later layers
- L20:294 suppresses
- L17:2264 suppresses
- L17:1051 excites
- L18:630 suppresses
- L17:2962 excites
- L26:37 suppresses
- L26:732 suppresses
- L19:221 suppresses
- L18:668 weakly suppresses
- L27:89 weakly suppresses
- L19:753 weakly suppresses
- L17:370 weakly suppresses
- L17:1635 weakly excites
- L26:70 weakly suppresses
- L17:404 weakly excites
- L24:226 weakly suppresses
- L17:544 weakly suppresses
- L26:55 weakly suppresses
- L19:299 weakly suppresses
- L22:90 weakly suppresses
- L19:886 weakly excites the space before a rounded number
- L23:21 weakly suppresses
- L18:2544 weakly suppresses
- L27:2529 weakly excites
Activation histogramA chart of how often a neuron takes each activation value.
How often this neuron fires at each strength. Almost always it does close to nothing (the tall bars in the middle); the snippets above are the rare moments it fired hard, amber for positive and blue for negative.
Logit lensReading which output tokens a neuron pushes toward or away from, straight off its output weights, before the later layers have their say.
The words this neuron nudges the model toward (amber) or away from (blue) when it fires. A hint, not a label: many neurons work through later layersOne step in the stack the text passes through. Qwen3-0.6B has 28 layers, each with 3,072 neurons. instead.
Promotes
- ロン
- 的母亲
- athe
- 晚报
- RN
- ată
- Bain
- rin
- ativas
- mant
Suppresses
- ..."
- ucing
- station
- azor
- 零
- star
- ";⏎
- ";
- istical
- omorphic