the letter n opening a word
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. 27 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 letter 'n' when it starts a word piece, often in Sundanese, Irish or other less common languages, and in words like 'navel' or 'nitrates'. Turning it up made the text shift to phrases like 'front door', and turning it down made it repeat 'living'. Neither change clearly relates to the letter.
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.
-
áil go slán. Ní dhéanann muid sonraí cártaí creidmheasa a stóráil agus níl aon rochtain againn ar shonraí a bh
-
ka anjeuna?|⏎|How difficult is it?: kumaha sesah na ieu?⏎|Can I help you?: tiasa abdi nulung anjeun?|⏎|Can you help me?: tiasa
-
When can we meet?: timana tiasa anjeuna sadaya papanggihan?|⏎|Who is knocking at the door?: saha na ngetuk panto?|⏎Now that you have explored questions in Sundanese
-
very good way to learn a language or start a conversation.⏎|What is this called?: ieu naon nyarios na|⏎|Why is it expensive?: naha ieu awis⏎|How would you like to pay?: kum
-
giant pandas, hsing hsing, ling ling, mainland china, mei xiang, mother and baby, mountain forests, national zoo, pandas, president nixon, temperate zone, time today<|endoftext|>Coffee is a plant (C
-
Early diagnosis and treatment are essential for a successful outcome.⏎Pain Near the Navel⏎One of the most common signs of appendicitis is pain near the navel. This pain often starts as a dull ache that gradually gets worse. In
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's hardest part of living as a living cat only living live live live live live live live live live live live live live live live live live live live live live live live live live live live live live live live live live live live
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
intentional injuries)⏎- Stroke (cerebrovascular diseases)⏎- Alzheimer’s disease⏎- Influenza and pneumonia⏎- Kidney disease (nephritis, nephrotic syndrome, and nephrosis)⏎Minister of Health
-
security on farms⏎Biosecurity is about reducing the risk of disease transmission. We know that bTB can be transmitted by direct contact (e.g. nose-to-nose) or indirect contact (e.g. via contaminated pasture) but simple exclusion
-
principally showed that meat and dairy are among the largest contributors to GHGE; whereas high intake of fruits, vegetables (F&V), and legumes/pulses/nuts consumption is associated with the lowest GHGE
-
organisms by oxidation! Hydrogen peroxide is considered the world’s safest all-natural effective sanitizer. It kills microorganisms by oxidizing them, which can be
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
- L22:2299 excites
- L21:2959 excites
- L23:1532 suppresses
- L19:595 excites
- L17:391 excites
- L26:1564 excites
- L21:567 suppresses the word 'in' after a verb or noun
- L8:14 excites
- L20:212 weakly excites
- L8:16 weakly suppresses
- L16:55 weakly excites
- L7:233 weakly suppresses
- L5:84 weakly suppresses
- L10:431 weakly excites
- L8:7 weakly excites
- L5:499 weakly excites
- L18:127 weakly excites
- L7:64 weakly suppresses
- L8:12 weakly excites
- L18:2052 weakly excites
- L8:10 weakly excites
- L18:569 weakly excites
- L6:8 weakly suppresses
- L5:110 weakly suppresses
Feeds later layers
None: this is the last layer.
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
- .Builder
- ANCEL
- bubb
- TEMPL
- дела
- atable
- ใหม
- ABCDEFGHI
- caps
- Corona
Suppresses
- onz
- asures
- ltk
- oreferrer
- .setDescription
- ovy
- ests
- ested
- atural
- .slide