"a" right after "as"
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 fires on the word "a" when it follows "as", as in "as a resurgent power", "as a tool" or "as a second language". Turning it up made the model repeat phrases, which doesn't clearly relate to this habit.
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.
-
The GIUK Gap forms the principal choke point between Russia’s great Northern Fleet and its strategic interests in the North Atlantic and all points south. The Russians, as a resurgent power, have modernized their military forces, but they still face the
-
contain redesigned genetic materials. Furthermore, they also give rise to redesigned biological systems, as a whole population/species is being altered or eliminated, thus changing the related ecosystems.
-
.)⏎In the dark all cats are gray. All persons are undistinguished till they have made a name.⏎Not room to swing a cat. Swinging cats as a mark for sportsmen was at one time a favourite amusement. There were several varieties
-
world’s largest and popular beaver dam, the Hover Dam at 1224 feet.⏎Dams are made with mud and stones by beavers as a protection against predators like wolves, bears and
-
Tutors and Students.⏎Biles, James J.; And Others⏎The guide is the second in a two-volume series designed to provide volunteer tutors of English as a Second Language (ESL) with skills and knowledge to individualize curricula and
-
observe it, we can learn so much” is the underlying philosophy of a number of innovative programmes being carried out in Uruguayan schools that are using gardens as a teaching resource, explained Edith Moraes, director of the national Primary Education
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: trees, birds, or animals. But the same names. So in the sky. So as in the sky. So as in the sky. So as in the sky. So in the sky. So in the sky. So as in
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
arks – Interactive Story Map.⏎Download these Missouri geologic maps to learn more about the diverse geology in the state.⏎These maps and others can be purchased by visiting us at 111 Fairgrounds Road in Rolla
-
these means the wisdom of “Father” Rosenkreuz, one hundred and twenty years after the latter’s burial, until about 16<|endoftext|>Gypsies have shared a common faith with Hungarians for many centuries. They put their talents,
-
, and sometimes-tumultuous American history ensued. But it got me thinking about the origin of the word thanksgiving and the fact that peoples of all nations have gathered through the ages to give thanks.⏎We don’t need to wait for blessings
-
ager and Amy Perry’s second grade classes presented poems, stories, songs, and jokes all involving trees.<|endoftext|>Obesity is
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
- L3:654 excites article 'a' after 'here are'
- L2:1977 suppresses
- L2:1426 excites
- L0:733 weakly excites
- L4:710 weakly excites
- L4:363 weakly excites
- L2:631 weakly suppresses
- L0:82 weakly excites
- L1:694 weakly suppresses
- L0:1256 weakly suppresses
- L1:2591 weakly excites
- L3:2707 weakly excites
- L3:976 weakly excites
- L2:1972 weakly suppresses
- L3:1172 weakly suppresses
- L0:2344 weakly excites
- L4:1579 weakly excites
- L0:406 weakly suppresses
- L0:39 weakly excites 'the' right after 'On' in 'On the other hand'
- L0:954 weakly excites
- L1:1288 weakly suppresses
- L2:1770 weakly suppresses
- L0:608 weakly excites
- L1:1165 weakly suppresses
Feeds later layers
- L7:146 strongly excites
- L6:1406 strongly excites
- L8:688 strongly excites
- L6:542 strongly excites
- L10:1476 strongly excites
- L7:385 strongly excites
- L7:327 strongly suppresses
- L6:1849 strongly excites
- L7:2115 strongly excites
- L8:1238 strongly excites
- L6:1734 strongly excites
- L16:1418 suppresses
- L6:410 suppresses
- L10:1190 suppresses
- L7:263 excites
- L10:661 excites
- L7:374 excites
- L7:257 excites
- L6:1655 excites "a" or "an" after "such" or launch verbs
- L9:439 suppresses
- L6:281 excites
- L16:2403 excites
- L6:415 excites
- L11:2798 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
- 这群
- ();)
- elves
- .Names
- 在一起
- eds
- 这些人
- 各有
- 姐妹
- closures
Suppresses
- ascal
- ểm
- lical
- anism
- циальн
- itzer
- angelo
- cursor
- handleMessage
- ết