unclear: 'your' in contact boilerplate
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. 10 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. 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 words in website and notice boilerplate, most often 'your' in phrases like 'contact your local office' or 'contact your librarian'. The trigger is not very consistent, and the steering outputs did not show a clear theme.
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.
-
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 the Moon's orbit where it lies diametrically
-
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 PrAACtically Martin Luther King Jr Day⏎Looking
-
reasons why your holiday shipments may not arrive on time. COVID-19 disruptions, congested ports, semiconductor shortages, stretched trucking capacity — the list goes on and on.⏎
-
range of therapies includes the following diseases of the head: acoustic neuromas / vestibular schwannomas, meningiomas, angiomas - arteriovenous malformations (AVM), metastases, trigeminal neuralgia and
-
can however freely search the site and view the abstracts and keywords for each book and chapter.⏎If you think you should have access to this title, please contact your librarian.⏎To troubleshoot, please check our FAQs, and if you can't
-
abstracts and keywords for each book and chapter.⏎If you think you should have access to this title, please contact your librarian.<|endoftext|>Information on creating school gardens for outdoor learning, compiled by the National
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: the design of a similar system. I have been considering the process of decision-making and how it might be applied to different scenarios. 1. In the following: "A" has been designed to be a control that has been tested in a
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
extrasolar planets, half of all that have been discovered. One recent discovery was an Earth-like planet orbiting in the habitability zone of its star where liquid water
-
collision that took place 1.3 billion years ago. When it occurred, about three times the mass of the sun was converted to energy in a fraction of a second.⏎The discovery has been accepted for publication in
-
as a cold wind blows⏎through a dark, dark cave.⏎Those words are split across multiple pages. The pattern of rhyming two lines and ending with a dark, dark cave continues. This sets a
-
the annexation of Austria by Nazi Germany, the rebellious actions of Benito Mussolini’s Italy and the ongoing boycott of Uruguay.⏎When France was chosen to host the World Cup two years before the tournament, South American giants Uruguay and Argentina announced
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
- L8:1988 weakly excites
- L8:2969 weakly suppresses
- L9:2777 weakly excites
- L7:1735 weakly suppresses
- L9:2438 weakly excites
- L0:2356 weakly suppresses
- L8:1857 weakly excites
- L6:2982 weakly excites
- L8:2282 weakly suppresses
- L7:211 weakly suppresses
- L6:168 weakly excites
- L6:1381 weakly suppresses
- L9:2514 weakly excites
- L4:1962 weakly excites
- L7:479 weakly suppresses
- L8:2727 weakly suppresses
- L5:3045 weakly suppresses
- L0:951 weakly excites
- L9:1800 weakly suppresses 'hands' after 'wash your'
- L8:2393 weakly excites
- L9:855 weakly suppresses
- L5:655 weakly excites
- L6:605 weakly excites
- L2:2669 weakly suppresses
Feeds later layers
- L12:359 excites
- L12:2283 weakly suppresses
- L13:2735 weakly suppresses
- L14:545 weakly excites
- L11:1946 weakly excites
- L13:2210 weakly suppresses
- L16:2949 weakly excites
- L11:436 weakly suppresses
- L16:1510 weakly suppresses
- L12:1436 weakly suppresses
- L11:428 weakly suppresses
- L14:1716 weakly excites
- L21:2696 weakly suppresses
- L21:2890 weakly suppresses
- L12:888 weakly suppresses
- L12:2010 weakly suppresses
- L22:1015 weakly suppresses
- L13:2422 weakly excites
- L18:987 weakly suppresses
- L22:2111 weakly suppresses
- L13:1919 weakly excites
- L13:1528 weakly suppresses
- L12:2013 weakly excites
- L15:128 weakly suppresses
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
- edException
- admit
- 雳
- elta
- lie
- enie
- 녕
- entifier
- gie
Suppresses
- Meetings
- iox
- caret
- harmless
- terminates
- Ter
- .clearRect
- castle
- eras
- rad