"other" and contrast words
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. 16 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them 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 most on the word "other" when one thing is being set against another, as in "the other peaks" or "the other plants". Words like "lower" and "south" also trigger it when they mark a contrasting side. Turning it up did not bring out a clear theme in the continuations.
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.
-
, there is a rocky giant called Mt. San Gorgonio. It is 11,502 feet tall—a half a head higher than the other peaks flanking it in the San Bernardino range. By early spring, when most
-
pered by factors such as genetic differences, upbringing, and lifestyle choices. By studying identical twins, however, the researchers were able to control for genetics and limit the other factors, as the twins grew up in the same households and reported similar lifestyles.⏎
-
of the vertebrae. Nerves branch from the spinal cord, pass through openings (foraminae) between the vertebrae, and branch to the lower body.⏎Primary Medical Reviewer William H. Blahd,
-
a horticultural term—it’s sort of like “vegetable,” in that you can kind of call anything a vegetable—but palms are not at all like the other plants commonly referred to as trees. They don’t have wood, for one thing
-
. Some philosophers who reject this metaphysical view have swung in the other direction and rejected the idea of the self altogether. David Hume said that the
-
1.4 million light-years away from Earth, so far away that it floats freely in space, unperturbed by the Milky Way.⏎Unlike the other hobbit galaxies, Leo T includes both old and young stars. It also contains
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 have to be sure that I have a good example, like I have to have a good example, like I have to have a good example, like, for example, maybe, like, like, like, maybe, like, for
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
s, and no Back-Cover Texts. A copy of the license is included in the page “GNU Free Documentation License”.⏎The copyright and license notices on this page only apply to
-
Awards program combines knowledge of STEM concepts with hands on activities.<|endoftext|>The sheer amount of gold in Fort Knox was so great that Auric Goldfinger, the villain of the 1964 James Bond film which carries his name, knew he
-
am presenting it here because it has been proven to work on many projects of all shapes and sizes. At the very least, your story should contain all of the elements described in the template. The story template looks like this:⏎<|endoftext|>What is a
-
For thousands of years, rubies were considered the gemstones of love, energy, passion, and power. In a class of its own
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
- L11:875 strongly suppresses
- L10:1377 strongly suppresses
- L11:331 strongly excites
- L15:1848 strongly suppresses
- L4:73 strongly suppresses
- L10:1162 strongly excites
- L13:1234 strongly suppresses
- L9:662 strongly suppresses
- L10:120 strongly suppresses
- L4:49 strongly suppresses
- L11:1168 strongly excites
- L15:869 strongly excites
- L15:1611 strongly suppresses
- L5:225 excites
- L2:205 excites
- L15:888 excites
- L13:824 suppresses
- L9:24 excites "the" before a named organisation or title
- L7:155 excites
- L3:344 excites
- L6:426 suppresses
- L14:2498 suppresses
- L9:59 excites "a" meaning per (a day, a year)
- L9:2198 excites
Feeds later layers
- L18:1547 excites
- L17:1542 suppresses
- L18:641 excites
- L17:2190 excites 'the' starting an object phrase
- L17:2970 excites
- L17:1179 suppresses
- L17:188 suppresses 'a' before a noun phrase
- L17:1652 suppresses
- L17:30 excites
- L18:1084 suppresses
- L22:1522 excites
- L18:314 excites
- L17:2135 excites 'the' after 'of' and similar words
- L19:2227 excites
- L18:799 excites
- L19:2815 excites
- L21:614 excites
- L19:3016 excites
- L20:901 suppresses
- L17:1481 excites 'the' after 'some of' or 'most of'
- L21:1741 excites
- L19:1543 excites
- L19:2115 excites
- L18:171 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
- converse
- dens
- ost
- od
- oses
- adows
- ones
- oden
- 一種
- jaws
Suppresses
- ún
- ()>⏎
- pillar
- (QtGui
- stood
- batim
- .yang
- fighter
- ręcz
- urate