'are' after a plural subject
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. 23 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 'are' when it follows a plural noun, as in 'The next largest cities are' or 'Other country sites are'. It is a verb-agreement signal and does not show clearly in the text the model wrote when it was turned up.
Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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.
-
fact, the largest city in the state, Charleston (the state capital) is home to just 49,736 people. The next largest cities are Huntington (48,638
-
at certain times, all the digits are⏎consecutive. How many times like this are there between midnight⏎and 7 a.m.?⏎These practical challenges are all about making a 'tray' and covering it with paper.⏎My cousin
-
You can also select a web site from the following list:⏎Select the China site (in Chinese or English) for best site performance. Other MathWorks country sites are not optimized for visits from your location.⏎Contact your local office<|endoftext|>It’s Pr
-
You can also select a web site from the following list:⏎Select the China site (in Chinese or English) for best site performance. Other MathWorks country sites are not optimized for visits from your location.⏎Contact your local office<|endoftext|>The point in
-
view this ebook on an iPhone, iPad or Android mobile device you will need the Adobe Digital Editions app, or BlueFire Reader or Txtr app. These are free, too. For more details see this article.⏎|Size: ||3
-
discouraged by the seemingly complicated written structure. With Music Theory for Dummies, understanding music has never been easier!⏎Note: CD-ROM/DVD and other supplementary materials are not included as part of eBook file.<|endoftext|>is a former municipality in Akers
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: it can is a word that is it is it is it is it is it is it is it is it is it is it is it is it is it is it is it is it is it is it is it is it is it
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
. . . And Young,” by Lt. Gen. Harold G. Moore and war correspondent Joseph Galloway. The 2002 movie, “We Were Soldiers,” starring Mel Gibson, was based on the Moore-Galloway book
-
OTHEISM⏎WE BELIEVE that MONOTHEISM, (TOWHID) has
-
about the Python-list<|endoftext|>Jayne Davis: “I am not convinced in my area of research there are less females choosing to take their career in this direction.”⏎We spoke to Jay
-
Intelligence. The Director of the Central Intelligence Agency (CIA) is appointed by the President with the advice and consent of the Senate.⏎Who manages the CIA?⏎We are pleased to announce that Bill Burns has been sworn in as the new Director of
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
- L22:309 strongly excites
- L19:1317 strongly excites
- L21:468 strongly excites
- L20:1423 strongly suppresses
- L19:1422 excites
- L18:444 suppresses
- L19:1629 excites
- L18:1096 excites
- L21:2936 suppresses
- L8:253 suppresses
- L20:1799 excites
- L19:545 excites
- L2:149 suppresses
- L4:450 excites
- L16:341 suppresses
- L19:880 suppresses
- L2:326 suppresses
- L4:298 excites
- L21:262 excites
- L12:642 excites
- L17:945 suppresses
- L18:1686 excites
- L17:147 suppresses
- L10:661 suppresses
Feeds later layers
- L24:600 strongly suppresses
- L26:72 strongly suppresses 'they' or 'we' after where or who
- L24:401 strongly excites 'it' referring back to a noun
- L24:1080 strongly suppresses
- L25:586 strongly excites
- L24:241 strongly suppresses
- L27:650 strongly suppresses
- L24:120 excites
- L27:321 excites
- L27:500 excites
- L26:2947 suppresses
- L26:467 excites
- L27:461 excites
- L25:2055 suppresses
- L26:1097 excites
- L24:277 excites
- L26:1417 suppresses
- L26:93 excites
- L25:347 excites
- L27:577 suppresses
- L26:874 suppresses
- L27:1256 excites
- L25:108 suppresses
- L26:926 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
- $is
- is
- meets
- grows
- survives
- does
- was
- doesn
- was
- goes
Suppresses
- 都可以
- 都有
- 都能
- 都是
- 都要
- 都
- 均可
- 都非常
- 都很
- are