commas in citations and name lists
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. 9 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 commas that sit inside formal references and lists, such as in a citation's volume number, a journal date, or a name followed by a title. When it was turned up, the output became repetitive rather than showing a 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.
-
erry, India. You may make a digital copy or printout of this text for your personal, non-commercial use under the condition that you copy this document without modifications and in its
-
and Chia-Liang Chiang, "Purification Study of Pervious Concrete Pavbement," International Journal of Engineering and Technology vol. 5, no. 5, pp. 532-535,
-
une-system⏎- Kennedy, David O. “B Vitamins and the Brain: Mechanisms, Dose and Efficacy—A Review.” Nutrients, vol. 8, no. 2, Jan. 201
-
aim of this legislation is establishing tour operators' responsibility towards customers and providing remedies to them (Gibson, 2016).⏎1.2 Surface, sea and air transport law in relation to the carriage of passengers⏎Below explained are
-
anxieties are indicators of healthy curiosity and strong moral fiber, while others are a source of severe stress. Knowing which is which can help you to navigate your personal, professional, and intellectual life more effectively.⏎Charlie Kurth, assistant professor in philosophy
-
per kilo of bodyweight than do younger people. But that’s because their lean body mass is less.⏎In a book called Biomarkers, authors William Evans, Ph.D. and
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 number of U.S. dollars in circulation in the United States Mint Museum Collection since March 15,000,001,001,001,001 in 1918-
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
a Code of Practice Concerning the Employment of Young Persons in Licensed Premises which a number of organisations have signed up to. IBEC, the employers’ organisation, recommends that pupils in examination classes should not
-
hearing aids when they are not in use.⏎Keep replacement batteries and small aids away from children and pets.⏎Certain conditions, including age, illness, and genetics, may contribute to hearing loss. Over several generations, modern life has added a host
-
methodology and data sources used for the calculation of the European steel recycling rate have been once again reviewed independently by the energy and waste specialist consultancy, Eunomia, who confirm the credibility of APEAL’s methodology and data.⏎www.ape
-
Future Bioenergy Workforce:⏎- K-12 Education: The McCall Outdoor Science School (MOSS), and Facing the Future, a non-profit organization, develop education curricula on bioenergy and biofuels, which is tested
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
- L4:3 strongly excites commas after citation fields
- L5:16 excites
- L3:71 suppresses
- L2:66 excites
- L5:1469 suppresses
- L6:798 excites
- L1:86 excites
- L6:620 suppresses "and" in lists of three or more
- L0:1596 excites
- L0:146 suppresses
- L6:427 weakly suppresses
- L1:1464 weakly excites
- L1:2683 weakly excites
- L6:1811 weakly excites
- L8:70 weakly excites 'and' between stacked superlatives
- L7:1223 weakly suppresses "and" joining two decades or times
- L6:912 weakly suppresses
- L0:1001 weakly excites
- L4:19 weakly suppresses
- L4:363 weakly suppresses
- L4:25 weakly excites
- L3:421 weakly suppresses
- L6:2945 weakly suppresses
- L5:102 weakly excites
Feeds later layers
- L14:35 strongly excites
- L18:366 strongly excites comma inside numbers and dates
- L10:335 strongly excites commas in catalog, date, and list metadata
- L15:588 strongly excites comma after an introductory phrase
- L19:726 excites comma inside large numbers
- L23:184 excites
- L17:118 excites comma after a clause ending
- L16:240 excites comma after 'sometimes' or 'the truth'
- L19:383 excites commas after a clause or list item
- L16:846 excites comma opening an aside or explanation
- L17:2285 excites
- L11:154 excites commas after citation numbers
- L22:144 excites
- L20:1298 excites
- L16:241 suppresses
- L16:1082 excites comma linking two clauses after 'and'
- L19:1506 excites
- L17:116 suppresses
- L17:585 excites
- L15:727 excites
- L18:1270 excites
- L20:476 excites commas that lead into a continuation
- L17:2001 excites
- L19:1826 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
- 以及其他
- Hezbollah
- 以及
- および
- która
- ahkan
- +⏎⏎
- "⏎⏎
- кер
- 和支持
Suppresses
- on
- ,
- ira
- 后再
- rex
- Ix
- 在这
- I
- ось
- 对他们