herb and remedy names, word fragments
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. 25 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 pieces of words in herbal and natural remedy texts, such as the 'g' in ginseng and the 'k' in kava. It also fires on a scientist's name split into pieces. Its negative side appears in dental treatment text. Turning it up or down gave no 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.
-
was, how does ATP get out to the cells of the nerve fibers to taste something sweet, bitter or umami?" study co-author J. Kevin Foskett, a neuroscientist at the University of Pennsylvania, told LiveScience.⏎
-
heart attacks, hypertension and increased susceptibility to infection, it’s easy to understand how one herb can have an all-encompassing, tonic effect.⏎For example, ginseng, another adaptogen, not only reduces stress
-
When those mice were given a taste test, they could not taste anything sweet, bitter or umami.⏎"That was a eureka moment for us," Foskett said. "This ion channel is absolutely essential for releasing the ATP. If
-
-related hormones, but research shows it also improves stamina and relieves fatigue. Eleuthero, often called Siberian ginseng, improves stamina and enhances immune function, as do reishi mushrooms —
-
is found in combination herbal products used as a sedative for promoting calmness and relaxation. Other herbs contained in these products include German chamomile, hops, kava, skullcap, and valerian.⏎Passionflower was formerly approved
-
and personal-care products (soaps, shampoos, hair sprays, and nail polishes).⏎Phthalates are used widely in poly
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 value of a variable function (x) is the value of the function at x (x function of x), which is logical, since the function itself is the function of x, and the value of the function at x is the value
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
floss between your teeth, curling it in a C-shape around each tooth.⏎6. Consider dental treatments:⏎Orthodontic therapy such as braces or clear aligners may be appropriate if you have crooked or misaligned teeth. Dental
-
dental problems before they worsen. In addition, your dentist can advise you on enhancing your dental health and propose fluoride treatments and sealants. Read more Composite Fillings.⏎4. Limit sugary and acidic foods or drinks:⏎Sug
-
possible tooth root inflammation, the dentist often removes the pulp completely. Crown or tooth bridge are then secured longer.⏎How long does a root canal treatment take at
-
another goes across your bottom teeth. It is held in place by small pieces of metal or ceramic that are put on each tooth. These pieces can be clear or tooth-colored. But then the fun begins. The elastic ties that hold the wire to
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
- L22:2287 suppresses
- L16:506 weakly excites
- L18:296 weakly excites
- L2:1611 weakly suppresses
- L3:1675 weakly suppresses
- L3:1523 weakly suppresses
- L2:1553 weakly excites
- L12:1394 weakly suppresses
- L0:614 weakly excites
- L5:2341 weakly suppresses
- L6:2299 weakly suppresses
- L13:2893 weakly excites
- L21:38 weakly excites
- L24:8 weakly suppresses
- L18:2879 weakly excites
- L5:1660 weakly excites
- L14:1217 weakly suppresses
- L22:2365 weakly excites
- L9:295 weakly excites
- L12:2525 weakly suppresses
- L22:151 weakly suppresses
- L4:2105 weakly suppresses
- L24:105 weakly suppresses
- L2:1625 weakly suppresses
Feeds later layers
- L26:2172 weakly excites
- L26:2089 weakly suppresses
- L26:2968 weakly excites
- L27:1774 weakly excites
- L27:559 weakly suppresses
- L26:566 weakly excites
- L26:904 weakly excites
- L27:771 weakly excites
- L27:1020 weakly suppresses
- L26:269 weakly suppresses
- L27:835 weakly suppresses
- L26:2568 weakly suppresses
- L27:1840 weakly excites
- L27:723 weakly suppresses
- L26:2234 weakly suppresses
- L27:2709 weakly suppresses
- L26:2007 weakly suppresses
- L27:1969 weakly suppresses
- L27:2513 weakly excites
- L27:918 weakly excites
- L27:1100 weakly suppresses
- L26:1151 weakly excites
- L26:705 weakly excites
- L26:1990 weakly 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
- 大纲
- Wiley
- 断
- 冲
- aways
- 突破
- 外包
- ColumnInfo
- 灸
- freezes
Suppresses
- pling
- ordinates
- moid
- plings
- ュー
- rouch
- ording
- ctor
- setId
- aku