second word of publication names
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. 10 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. 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 lights up on the later part of a well-known name, such as the 'ica' in Britannica, 'Science' after 'Popular', or 'Privacy' after 'Electronic'. It seems to predict the rest of a magazine, organisation or publication name. Turning it up gave mostly unrelated text, with one output mentioning the New York Times.
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.
-
Children were more vulnerable to rotavirus (p = 0.008, OR<|endoftext|>A Global Index of Sustainable Well-Being⏎Source: The New Economics Foundation⏎From the Executive Summary:⏎There is a growing global consensus that we need
-
marsat, but was withdrawn in 2008 and replaced by the broadband global area network (BGAN), a global satellite Internet network.⏎Techopedia explains Regional Broadband Global Area Network (
-
the ore is crushed and agglomerated prior to heap leaching. High grade ores and ores resistant to cyanide leaching at coarse particle⏎Gold processing Britannica.com⏎The mining of alluvial deposits and, later, lode or
-
's] records unless specifically permitted by the Act.” See Your Right to Federal Records: Questions and Answers on the Freedom of Information Act and the Privacy Act. Electronic Privacy Information Center, 1992.⏎The American Inquisition: Federalism
-
The early period, 1853–1930⏎Our editors will review what you’ve submitted and determine whether to revise the article.Join Britannica's Publishing Partner Program and our community of experts to gain a global audience for your
-
they need to escape, but then they are too full to bend over and sting.⏎• In “The American Frugal Housewife,” (the Better Homes and Gardens of its day) the author recommends using ear wax as lip balm.<|endoftext|>
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 can't be as specific as the time of my own age. So, what is the time of my own age? Maybe the time of my own age. So, how I am at the time of the age of the person.
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
to Rapid City, fire weather meteorologists are watching conditions closely. And until we receive widespread, heavy moisture they'll be monitoring what is known as the Keetch-Byram Drought Index or KBDI.⏎It measures the amount of precipitation
-
of age or older completed a survey that included a food frequency questionnaire and cognitive assessment. Cognitive function was checked again during four assessments over 11 years using the Modified Mini-Mental State examination (3MS), which is graded on a 1
-
surely knew that this Act would shed light on the seditious acts of government officials. Curious.⏎Indeed, the Act itself, found in the Statutes at Large, 80 Stat. 383 stated:⏎"Except as
-
https://youtu.be/qlSxPouPCIM⏎The post 1929 Stock Market Crash and the Great Depression. first appeared on COMPLIANT PAPERS.⏎The post 1929 Stock Market Crash
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
- L7:1847 excites
- L7:2571 suppresses
- L9:171 suppresses
- L9:450 excites
- L9:50 weakly suppresses
- L3:1124 weakly suppresses
- L6:1866 weakly excites
- L3:455 weakly suppresses unclear: mid-name word fragments
- L7:180 weakly excites
- L7:2229 weakly excites
- L1:1530 weakly suppresses
- L0:868 weakly excites
- L0:2010 weakly excites
- L7:2174 weakly suppresses
- L6:2164 weakly suppresses second half of multi-word proper names
- L9:904 weakly excites
- L8:114 weakly suppresses
- L3:169 weakly suppresses
- L5:26 weakly excites
- L4:1454 weakly excites
- L7:2492 weakly excites
- L9:1473 weakly suppresses
- L3:809 weakly excites
- L2:1824 weakly excites
Feeds later layers
- L12:3020 excites
- L13:1215 excites
- L16:2217 excites
- L18:2292 excites
- L13:2755 excites
- L19:1519 excites
- L21:2846 excites 'at' after a job title, before an institution
- L14:1677 excites
- L20:2813 excites
- L19:2135 excites
- L21:2850 excites
- L13:2165 excites
- L19:2411 excites
- L16:2166 excites
- L16:2208 excites
- L21:2520 excites
- L12:139 suppresses
- L12:756 excites
- L11:457 excites
- L15:2628 excites
- L11:1182 excites
- L22:2134 excites
- L13:450 suppresses
- L18:3040 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
- 时报
- Sheet
- 报社
- #endregion
- 网站
- .microsoft
- Times
- chapel
- atoria
- Institute
Suppresses
- uant
- amiento
- uating
- ungan
- utches
- uate
- itious
- _own
- 환
- ung