opening parenthesis after a claim
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. 17 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 an opening bracket "(" that starts an aside, a citation or an explanation right after a statement. Turning it down made the model spam brackets and dashes, and turning it up made it write stray brackets and odd parenthetical asides.
Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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.
-
purchase of a lottery ticket cannot be rationally accounted for by decision models based on expected value maximization, because the expected value of a lottery ticket is always zero (or close to it). However, the non-monetary benefits obtained from playing the
-
yes/no” is perfectly adequate. This is not the case with a feature that has a specific tolerance i.e.: a nominal and an upper and lower spec limit (1.0 +/- .005). It might be tempting to put a
-
Charged Issue⏎One reason this is such attractive fodder for cable news shows is that the Waxman Markey bill is a politically-charged social and economic issue (which pretty much guarantees it will bring out strong opinions here, no matter
-
of discrimination? “Currently, the prison is the nation’s most racialized type of real estate (Robertson, 2006).” Many inmates must participate in the
-
will be particularly difficult when urbanization requires genetic adaptations, having implications for urban species diversity.⏎The rapid growth in urban development and increasing concentration of people in these areas (Cohen 2003; UN 20
-
carry one or two computers with us at all times. The present is called the Information Age precisely because our brains can delegate a lot of information storage and distribution indirectly (through our fingertips) to electronic devices. And it isn’t as if the possibilities
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: ( ( -1/2 -1/2 -1/2 -1/2 -1-1-1-1-1-1-1-1-1-1-1-1-1-1-1
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
Nurses Association - (NBNA) Welcome - Dedicated⏎advancement of African-American nurses and improving the healthcare
-
one eye sees one image and the other eye, the other. By having the two layered images at slighlty different ‘angles’, in effect, (special cameras need to be used to create these
-
conditions can also cause loss of appetite in seniors. These include:⏎- Dementia or Alzheimer’s disease. As dementia progresses, it’s common for both appetite and weight loss to occur.⏎
-
chures⏎- Market Trends⏎- Material Properties Charts⏎- List Rental⏎- Custom Content & Marketing Services⏎Imagine stepping outside for a breath of fresh air-and encountering 3 billion tons of air pollution. According to the Manufacturers of Emission
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
- L3:57 strongly excites
- L16:264 suppresses opening parenthesis in citations and asides
- L4:64 excites
- L10:66 suppresses opening parenthesis after a citation or term
- L6:468 excites
- L0:1622 excites
- L0:11 excites
- L11:50 excites
- L8:770 excites
- L1:102 excites
- L4:45 suppresses
- L5:275 excites
- L10:1041 suppresses opening parenthesis before phone or code
- L6:547 suppresses
- L0:1252 suppresses
- L8:354 excites opening parenthesis before a citation year
- L16:1143 suppresses
- L6:13 excites
- L5:330 excites opening parenthesis after a noun
- L15:82 weakly suppresses
- L10:142 weakly suppresses opening parenthesis mid-sentence
- L16:428 weakly excites unclear: mid-sentence commas, 'most' before a verb
- L5:2795 weakly excites
- L15:114 weakly excites
Feeds later layers
- L19:453 strongly excites
- L20:2173 strongly excites
- L18:1501 strongly suppresses
- L25:360 suppresses
- L23:224 suppresses
- L20:104 suppresses
- L18:721 excites opening parenthesis after a word
- L20:521 suppresses
- L18:980 excites opening parenthesis after a noun phrase
- L24:68 excites
- L19:2251 excites opening parenthesis after a name
- L26:13 excites
- L20:1451 excites
- L20:1841 excites opening parenthesis before a citation
- L26:234 suppresses
- L27:218 suppresses
- L18:682 excites
- L20:758 suppresses
- L27:1463 excites
- L19:1864 excites
- L26:88 excites
- L24:102 suppresses
- L19:889 excites
- L18:1247 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
- )::
- .*)
- .*?)
- );⏎
- ,),⏎
- ,),
- 惩戒
- ); ⏎
- culus
- );⏎⏎
Suppresses
- "](
- (
- `](
- (
- (
- (ib
- (
- "(
- Gle
- '](