'of' in fixed phrases like table of contents
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. 12 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 fires on the word 'of' in set phrases such as 'Table of Contents', 'range of movement', 'field of view', and 'proof of concept'. It fires weakly or not at all on other words. Turning it up did not make the writing about any single topic.
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.
-
plant care tips to help you efficiently care for and grow this plant’s root system.⏎Table of Contents⏎The family of Medusa Head Plants contains more than three hundred varieties of
-
by University of Minnesota students. You can see all of the details involved in making it in the video below.<|endoftext|>Create a Graph⏎Library Home || Full Table of Contents || Suggest a Link || Library Help⏎|National Center for Education Statistics
-
and uncover the remarkable ways in which they assist us in diagnosing and managing our health.⏎Table of Contents⏎Cancer is a disease that affects millions of people worldwide. It is
-
<|endoftext|>FROM ISSUE NUMBER 15 ~ SPRING 2013 GO TO TABLE OF CONTENTS⏎The Art of Limited Government⏎In his first inaugural address, Thomas
-
you how to weld pipe correctly. By the end, you will be able to start preparing as a pipe welder and having joints that pass stress tests.⏎Table of Contents⏎- 1 How welding works⏎- 2 Which welding is best
-
time during a comets apparition even⏎binoculars will be too powerful...⏎A telescope is a big mistake, it will have much too narrow a field of view.⏎Please, you should not just go out to an unfamiliar rural "dark
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: what is the value of the expression $f(f(f(f(x)))$ evaluated at x? when the input is processed through four different functions, and then the output is processed again through four different functions, and the expression is evaluated again at
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
ask a parent or adult before you fill out any forms, or give out money or credit card details.⏎- Don't accept any offers that seem too good to be true - they probably are.⏎Remember to always be
-
when you're online to keep your identity safe.⏎- Only Make Purchases On Trusted Sites⏎When deals seem too good to be true, they just may be - you might be paying as a victim of identity
-
Fukushima, ...) and model assessment using single effect tests or integrated experiments (QUENCH, PARAMETER, ...).⏎In the SACRE group SCDAP/RELAP5 is used for core degradation calculations and model assessment using single effect tests or integrated
-
schools must meet graduation goals and middle and elementary schools must meet a third academic indicator. Schools and districts that miss goals are identified for improvement.⏎Visit the Adequate Yearly Progress Page⏎Federal law requires states, school districts, and schools to produce
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
- L11:393 strongly excites
- L10:1705 strongly excites
- L9:746 strongly suppresses
- L8:1480 strongly suppresses
- L10:2514 strongly suppresses
- L10:952 strongly excites
- L7:2370 strongly suppresses
- L6:2164 strongly suppresses second half of multi-word proper names
- L10:2510 strongly suppresses
- L8:849 strongly suppresses
- L10:2105 strongly suppresses
- L8:1976 strongly suppresses
- L11:1384 strongly suppresses
- L9:1276 strongly excites
- L9:845 strongly excites
- L10:129 strongly suppresses
- L10:1784 excites
- L9:2613 suppresses
- L8:2388 suppresses
- L10:2035 suppresses
- L8:852 suppresses comma inside large numbers
- L7:1684 excites
- L9:716 excites
- L9:2392 suppresses
Feeds later layers
- L17:1042 strongly excites
- L15:646 strongly excites
- L13:1929 strongly excites
- L13:450 strongly suppresses
- L19:2411 strongly excites
- L20:1292 strongly excites
- L14:1403 strongly excites
- L18:558 excites
- L19:1432 excites
- L15:966 excites
- L15:1964 excites
- L19:107 suppresses
- L16:2117 excites
- L19:1385 excites
- L18:151 suppresses
- L18:819 excites
- L20:118 excites
- L17:533 excites
- L13:738 excites
- L14:528 excites
- L20:452 suppresses
- L18:708 excites
- L18:1165 excites
- L15:227 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
- ycastle
- .force
- .),
- .").
- PCS
- 纣
- 商会
- Merc
- бу
- .Factory
Suppresses
- ting
- agnostics
- uated
- irmed
- above
- hoot
- NING
- 走
- ning
- ging