the 'D' in vitamin D
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 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 letter D when it follows the word vitamin, as in vitamin D. It hardly reacts to anything else. When it was turned up, the model's replies did not clearly change in any way we could read.
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.
-
that are linked with depression.⏎Vitamin D Deficiency⏎According to studies reported by MayoClinic.com, people who are
-
their diets. Women in particular are prone to iron loss due to menstruation since they lose some of their body’s iron in blood and other discharge.⏎Vitamin D, like iron and calcium, has many uses and old age does not make it
-
butter, cheese, cream and fortified milk; oysters; and certain other fortified foods, such as margarine, soy milk and some breakfast cereals. Vitamin D is also available in dietary supplement form.⏎Because Vitamin D promotes your body’s absorption
-
in. In addition, because the industry is not well regulated, there is no real guarantee that the ingredients and dosage on the label will be accurate.⏎Vitamin D promotes calcium absorption in the body, and getting enough is critical to health and well
-
,000 mg per day⏎1 to 3 years old: 700 mg per day⏎Your diet should also include adequate amounts of vitamin D. Vitamin D is essential for the body to absorb calcium. A lack of vitamin
-
less sunny locations. A recent review reported that many studies found reduced number of dental cavities for those with vitamin D supplementation.⏎There is also good evidence that vitamin D reduces risk of sepsis (blood infection). This infection often sends people to
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: you have to be very careful to have a good time, and that you need to be careful to be safe. So, according to this, I need to be very careful to be safe. So, based on this, I need to
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
exceptional athletes living with Parkinson's.⏎Read on for all of November's Parkinson's news!⏎SCIENCE AND RESEARCH Updates⏎- Chu et al. found dopaminergic neurodegeneration can independently occur of alpha-synuclein aggregation. They
-
…and the culprit is poor nutrition.⏎The “hunger, disease and ignorance” Mandela spoke of are often almost inevitable bedfellows in many of the Early Childhood Development (ECD) Centres and schools in South Africa’s less affluent communities and
-
mean that you don’t have enough insulin in your body. You may need to check for ketones when:⏎Your blood sugar is above 250 mg/dL (milligrams per deciliter
-
MHIRT program through San Diego State University and was published in the Journal of Ethnopharmacology.⏎Per
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
- L2:110 strongly excites
- L1:773 strongly excites
- L0:233 excites
- L4:1373 suppresses
- L1:10 excites
- L0:553 suppresses second half of a split name
- L0:1561 weakly suppresses
- L2:1807 weakly excites
- L2:558 weakly excites
- L0:959 weakly excites
- L4:84 weakly suppresses
- L2:2569 weakly excites
- L3:811 weakly excites
- L3:1605 weakly suppresses
- L2:1942 weakly excites
- L2:1918 weakly suppresses
- L3:1751 weakly suppresses
- L1:1561 weakly excites
- L4:2484 weakly excites
- L2:2694 weakly suppresses
- L3:1929 weakly excites
- L2:2192 weakly excites
- L4:1360 weakly excites
- L1:1493 weakly excites
Feeds later layers
- L18:105 strongly suppresses
- L17:300 strongly suppresses
- L20:200 strongly suppresses
- L16:572 strongly suppresses
- L10:180 strongly suppresses
- L22:473 strongly suppresses
- L19:372 strongly suppresses
- L18:1466 strongly suppresses
- L21:1912 strongly suppresses
- L21:2743 strongly suppresses
- L19:526 suppresses
- L26:581 suppresses
- L23:2947 suppresses
- L27:2376 suppresses capital D starting a split word
- L27:2889 suppresses
- L24:2782 suppresses
- L22:2579 suppresses
- L22:1511 suppresses
- L23:400 suppresses
- L26:3052 suppresses
- L27:2799 suppresses the letter d inside words
- L18:1493 excites
- L17:2511 suppresses
- L27:2968 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
- brook
- downs
- RedirectTo
- eam
- 동
- ��
- maf
- due
- ups
- 渑
Suppresses
- 早晨
- ʑ
- 早上
- 伙
- 晚上
- avour
- olor
- fastest
- exter
- 中午