"one" as a pronoun after a verb
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. 3 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 word "one" when it stands in for a noun, as in "getting one" or "find one of the few". Turning it up made the text loop and repeat, so the steering does not show much.
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.
-
have a word processing or spreadsheet program for your computer, consider getting one to help you be responsible in school. Once you learn how to use the program
-
izations, and memories with a computer. An Imagination Interface, perhaps. By any name, it doesn’t take much time to realize that you have always wanted one. Think of a world where you could selectively download your life’s memories- sights
-
as cancer and heart disease? Next best is salad, then fruit.⏎Remember that if you have a health insurance or life insurance policy, or are thinking about taking one out, good health can help to lower your premiums - another reason to reach for
-
.⏎It can quickly seem like nothing we do matters all that much. "What would you do if you were stuck in one place, and every day was exactly the same, and nothing that you did mattered
-
so rare that generations of men and women never live to see one, and so significant it enabled astronomers armed with telescopes and trigonometry to
-
seen. This giant rat also seemed very tame and approachable towards the scientist. If you’re not afraid of rats then you’re probably asking yourself where to get one as a pet? Well, this extremely rare animal can only be found in one
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 first time I saw the light of the world in a new way. But that's not to be gone. What's interesting in the first place is the way to the north of the mind. But that's not to be gone.
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
to guide the treatment of children and adolescents with schizophrenia.⏎“Although the U.S. Food and Drug Administration (FDA) has recently approved the use of aripiprazole and risperidone for adolescents with schizophrenia, few controlled data are
-
such as olanzapine (Zyprexa), risperidone (Risperdal), ziprasidone (Geodon), or aripiprazole (Abilify) might have less side effects than haloperid
-
of the explosive eruption of Hecates Tholus provides more evidence of recent Mars rumblings. In December, members of the same research team revealed that calderas on five major Mars volcanoes were repeatedly active as little as 2 million
-
iusLeadership | Messenger birds | Lugh Long-arm | Odin | Rigveda | Thor | Odin and Tyr | Balder myth | Trinity | AS-ideas |Goddesses | indexBR | index
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
- L2:70 weakly excites
- L0:396 weakly excites
- L2:2917 weakly suppresses unclear: 'ardu' and 'axi' word fragments
- L1:2267 weakly suppresses
- L0:156 weakly suppresses
- L2:2545 weakly suppresses
- L0:2989 weakly suppresses
- L0:1609 weakly suppresses
- L1:778 weakly suppresses
- L0:1071 weakly excites
- L0:1 weakly suppresses
- L2:1494 weakly suppresses
- L0:2759 weakly excites
- L0:1301 weakly suppresses
- L2:506 weakly suppresses
- L2:1160 weakly excites
- L1:2198 weakly excites
- L0:95 weakly suppresses
- L2:1430 weakly suppresses
- L2:1213 weakly excites
- L2:2328 weakly suppresses
- L1:94 weakly suppresses
- L0:312 weakly suppresses
- L2:1786 weakly excites
Feeds later layers
- L4:323 excites
- L5:860 weakly excites
- L10:1875 weakly excites
- L19:2967 weakly suppresses
- L4:2589 weakly suppresses
- L12:2888 weakly suppresses
- L7:1109 weakly excites
- L25:2429 weakly excites
- L8:1873 weakly excites
- L10:1278 weakly excites
- L15:1065 weakly excites
- L9:802 weakly suppresses
- L16:2822 weakly suppresses
- L17:2482 weakly suppresses
- L19:2804 weakly suppresses 'the' after 'in' or 'use'
- L12:2637 weakly suppresses
- L24:653 weakly suppresses
- L12:1582 weakly excites
- L8:1776 weakly excites
- L16:297 weakly suppresses
- L25:2246 weakly suppresses
- L9:679 weakly excites
- L18:2909 weakly excites
- L23:1867 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
- ax
- Ax
- 有限公司
- sew
- preached
- ester
- ций
- spar
- 出现
- all
Suppresses
- EMON
- atoire
- Berm
- itemap
- körper
- cano
- 金字塔
- usat
- bove
- urname