unclear: weak, scattered word hits
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. 5 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 responds only weakly, and the words it picks out have no clear shared theme. Its steering effects did not point to any particular topic, so we cannot say much about what it does.
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.
-
pit, the King would not be saved nor would be cured. The people were surprised, since they believed that he was dead. They went to the pit and dropped a rope
-
reacting in a certain situation. The question represents a “mismatch ” between the potentially necessary experience and the experience, which is available. The mismatch is used for controlling activity.⏎2. Each question that any pupil has should be casual and be generated
-
the palm of your right hand and try to squeeze it as hard as you can. Now, place three fingers from your left hand in the palm of your right hand and squeeze. Notice how much easier it is to squeeze by adding
-
at addressing climate and sustainability issues in their community.⏎These young people are green entrepreneurs and they are all developing cool projects aimed at addressing climate and sustainability issues in their community.⏎For instance, Syuichi Ishibashi from Japan has developed an impressive
-
,⏎South Africa. [photo Trevor Hardaker ©]⏎Southern black tit female, Thornybush Game⏎Reserve, South Africa. [photo Trevor Hardaker ©]⏎Distribution and habitat⏎Occurs from Zambia and southern Angola to southern Africa.⏎Here
-
occur in anyone, of any gender, at any age.⏎Eating disorders most commonly affect women and occur in adolescence and young adulthood, though they can occur in anyone, of any gender, at any age.”⏎Eating disorders often occur alongside other
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: "I do not know what this is, but I have been thinking about it. So I do not know what this is. I do not know what this is. I do not know what this is. I do not know what this
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
the day, to help prevent constipation.⏎Certain groups of people are at a higher risk for calcium deficiency. These groups include post-menopausal women, amenorrheic woman (no menstrual periods), people with lactose intolerance and vegetarians
-
Researchers believe that fibromyalgia amplifies painful sensations by affecting the way your brain and spinal cord process painful and nonpainful signals.⏎Fibromyalgia is commonly thought of as a condition that affects adults. However, fibromyal
-
through our world when two black holes bash together in a distant galaxy are on the order of a billionth of a billionth the diameter of an atom. LIGO detects them by proxy, using high powered lasers to measure tiny changes in the distance
-
was a victory for the U.S. in the Cold War Space Race with the Soviet Union who, in 1961, had sent Yuri Gagarin to the Cosmos to become the first human in outer space and the first to orbit
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
- L2:433 strongly suppresses
- L2:234 strongly excites
- L2:33 strongly excites
- L4:246 strongly suppresses
- L2:727 excites
- L1:46 excites
- L4:418 suppresses
- L3:403 suppresses
- L3:757 suppresses
- L4:41 excites
- L3:16 excites
- L4:100 suppresses
- L4:446 excites
- L2:334 suppresses
- L4:278 suppresses
- L3:404 weakly excites
- L2:1630 weakly suppresses
- L3:429 weakly excites
- L1:1059 weakly excites
- L3:365 weakly excites
- L3:59 weakly suppresses
- L4:535 weakly excites
- L3:102 weakly excites
- L3:421 weakly excites
Feeds later layers
- L9:1574 strongly excites
- L10:1475 strongly excites
- L9:733 strongly excites
- L9:1020 strongly excites
- L12:138 strongly excites
- L9:421 strongly suppresses
- L18:998 strongly excites
- L10:181 strongly excites
- L13:1168 strongly excites
- L18:942 excites
- L26:486 suppresses
- L14:1106 excites
- L8:56 excites
- L17:331 excites
- L10:415 excites
- L8:630 suppresses
- L13:230 suppresses
- L16:776 suppresses
- L18:688 excites
- L19:77 suppresses
- L19:140 excites
- L15:2228 excites
- L15:1550 excites
- L16:643 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
- */}⏎
- `,⏎
- ⏎⏎⏎
- ];⏎⏎⏎
- .*/⏎
- '.$
- ersh
- `;⏎
- brig
- .`,⏎
Suppresses
- ... ⏎
- ) ⏎
- " ⏎
- """ ⏎
- ⏎
- ... ⏎
- " ⏎
- % ⏎
- perfil
- swana