'in' meaning 'in this case' or 'in between'
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 'in' when it means something like 'in this case', 'in between' or 'in terms of', as in 'anything in between' or 'in that'. When it was turned down, the model got stuck repeating 'in in in'. Turned up, the writing drifted to unrelated topics.
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.
-
spirits can happen outside. It just requires that we embrace a broader vision of what “outside” means: an urban playground, a remote national park, and everywhere in between. We need to Rethink Outside and center equity when doing so.
-
ards have been going in and out of style for years. What is the reason for this changing style of hair and no hair? We’re going to step back in time to the beginning and examine the history of the beard.⏎In prehistoric times
-
Engage in Reading Adventures⏎There is no better time to read than during the dog days of summer. Whether it’s sci-fi, historical fiction, or anything in between, students can be encouraged to find things that they are excited to explore during
-
capital of the South,” a Jet Age airport promised to put what had once been a minor railroad town within a day’s reach of London and Paris – and, in the minds of some Atlantans, on the same plane of international significance as
-
different types of treatment<|endoftext|>The concept of a graph has been around since the dawn of mechanical computing and for many decades prior in the domain of pure mathematics. Due in large part to this golden age of databases, graphs are becoming increasingly popular in software
-
,” the phenomenon captured by the telescope is what’s known as a reflection nebula – a piece of debris left behind by the formation of a newborn star, or in this case, a small, multiple star system known like V380 Orion
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: in I have, on I have, I have, I have on I have on on I have, I have on on again, on I have, I have, I have, I have on I have, on I have, I
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
Comparative Analysis⏎- 人文学院-已发表论文 ⏎“助动词”在汉语中具有较大的作用,使用率高,但是泰国学生经常在学习和使用汉语助动词中产生各种的偏误。本文通过
-
and Huntington’s, are characterized by protein aggregation (clumping) in the nerves’ cells, impairing their function.⏎Normally, when a protein is made in the body, it is folded into the 3D shape it needs to perform its
-
kind, including, without limitation, representations and warranties as to the functionality of the translation feature or the accuracy or completeness of the translations.⏎Translations are not retained in our system. Your use of this feature and the translations is subject
-
to be an accessible topic for middle school students.19<|endoftext|>毋忘在莒(Pinyin: Wu Wang Zai Ju) is a famous Chinese idiom
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
- L4:289 strongly excites
- L3:493 strongly suppresses
- L0:781 suppresses
- L1:1179 suppresses 'in' after 'result' (results in)
- L5:490 excites
- L9:1025 excites
- L8:351 suppresses 'in' after 'high' or 'dressed'
- L10:262 suppresses
- L14:1625 excites the word 'in' after a noun
- L0:1280 excites
- L5:729 suppresses
- L4:1966 suppresses
- L8:3068 suppresses
- L2:2307 excites
- L15:871 suppresses
- L6:846 suppresses
- L15:1328 weakly suppresses
- L15:78 weakly excites
- L2:1463 weakly suppresses
- L14:1389 weakly suppresses
- L11:2856 weakly suppresses
- L1:1922 weakly suppresses
- L15:115 weakly excites
- L14:1122 weakly suppresses
Feeds later layers
- L17:706 excites
- L24:893 suppresses
- L21:1935 excites
- L19:704 excites
- L17:2619 excites
- L23:1911 excites
- L22:535 excites
- L18:869 excites
- L20:1574 excites
- L24:2163 excites
- L25:2032 excites
- L19:1345 excites
- L27:168 excites
- L25:1865 excites 'in' in 'keep in mind'
- L23:1366 excites
- L17:2821 excites
- L26:825 suppresses
- L24:2051 excites
- L25:2957 excites 'in' starting a word fragment
- L27:405 suppresses
- L24:2134 excites
- L18:2615 excites
- L18:698 excites
- L18:2832 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
- variants
- form
- var
- asm
- emple
- /out
- StateException
- variant
- terms
- finity
Suppresses
- 在外面
- сайте
- itur
- ально
- 在学校
- 在路上
- 在国内
- 在国外
- 在生活中
- 在深圳