line break after list items or paragraphs
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. 18 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 line break that comes right after a sentence or list item, where the next line is likely to start a new bullet or paragraph. When it was turned up, the text drifted into rambling, repetitive phrases with no clear topic.
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 best way to take their messages to the wider community by producing ads or other media about the topic.⏎- Have learners create social media campaigns around a particular issue⏎- Develop a public
-
Age And Address And Drawing A Self Portrait And A Picture Of Where They Live Alphabet This Early Writing Worksheet Gives Your Child Practice Writing The Entire Alphabet In Lowercase Letters⏎Kindergarten Reading Writing Printable Worksheets⏎Kindergarten<|endoftext|>- as in the cold war
-
qual<|endoftext|>Why did Non-Cooperation Movement gradually slow down in cities? Explain two reasons.⏎The Non-Cooperation Movement in the cities gradually slowed down for various reasons⏎
-
to eating? Try these tips to get started down the path to health and longevity!⏎- Aim for at least one fully vegetarian meal a week and go from there⏎If you are new to this concept it may be a little intimidating to completely up
-
what has happened with similar initiatives in other areas (both successes and failures), and shifts in political priorities and policy intentions. Keeping a diary of events can be useful⏎- Build in a periodic check on these kinds of issues into the agenda of your
-
single unit. This may alleviate pain in a motion segment.⏎Spinal fusion, however, has well known potential disadvantages, including:⏎- Loss of motion and flexibility⏎- Permanently altered
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 a number of people in the age of 13 and 15, and some of the age groups are quite old in the age of 13 and 15, and this is a good age for some of
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
- Pupils need access to technology and research materials to produce PSAs or social media campaigns⏎- This resource is only available on an unencrypted HTTP website. It should be fine for general use, but don’t use it to share any
-
in a circle⏎(C) 2009 Copyright Math Open Reference. All rights reserved<|endoftext|>Use chain rule⏎F'(x) = dF(x
-
programmers⏎Verbal – masters of word manipulation, they can be quite persuasive. Their command of language gives then a love for storytelling, & a tendency to arguing. These students can have careers in marketing or politics⏎Visual – creating vivid mental pictures
-
the base if cut back by fire[⏎The nectar is used to make a sweet drink[⏎The roots and bark are antipyretic and antiseptic. They are used traditionally to treat headaches[⏎If you have any useful information about this
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
- L15:2086 strongly suppresses
- L14:2593 strongly suppresses
- L16:1047 strongly excites
- L15:2653 excites
- L14:2364 excites
- L15:2717 excites
- L16:489 suppresses
- L17:1048 suppresses
- L17:1772 suppresses
- L9:1891 excites
- L16:1123 excites
- L17:1213 excites parenthesis opening a side note
- L16:432 suppresses period after month or initial
- L17:642 suppresses
- L17:659 suppresses
- L14:853 suppresses
- L8:740 excites
- L17:1342 excites
- L14:2130 suppresses
- L9:2447 excites
- L15:345 excites
- L17:856 excites
- L15:1207 excites
- L16:2398 excites
Feeds later layers
- L19:389 strongly suppresses
- L20:491 strongly excites
- L21:102 strongly suppresses
- L21:233 strongly excites
- L25:40 strongly suppresses
- L25:297 strongly excites
- L23:171 strongly excites
- L26:11 strongly excites
- L21:759 strongly excites
- L27:1801 strongly excites
- L23:444 strongly excites
- L21:194 strongly suppresses
- L22:282 excites period then line break in lists
- L21:147 excites
- L26:1018 excites
- L27:385 suppresses
- L24:507 suppresses
- L25:940 suppresses
- L24:12 excites
- L27:319 suppresses
- L20:699 suppresses
- L24:145 excites
- L27:1053 excites
- L27:703 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
- ⏎⏎
- ⏎⏎⏎
- ⏎
- ⏎⏎⏎⏎
- ⏎⏎⏎⏎⏎⏎
- ⏎ ⏎
- "⏎⏎
- ”⏎⏎
- `⏎⏎
- )⏎⏎
Suppresses
- .', ⏎
- .',⏎
- .');⏎
- .'); ⏎
- .",⏎
- ."); ⏎
- .';⏎
- .").
- .'));⏎
- .";⏎