'young' before people or adults
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. 24 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 lights up on the word "young" when it comes right before "people" or "adults", as in "young people" or "youth and young adults". When it was turned up, the model's writing drifted into unrelated topics, so the steering does not show a clear effect.
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.
-
them.⏎But they clearly contain nicotine, which is addictive. Public health advocates warn that e-cigarette makers are using the same strategies that tobacco companies employed to attract young people — candy flavorings, for example. Lax state-level regulations also make
-
unfortunately, in the absence of evidence-based public health messages regarding the health risks of e-cigarettes, television advertising is promoting beliefs and behaviors that pose harm to youth and young adults and raise public health concerns.”⏎Researchers analyzed TV advertising data and viewership ratings
-
simply believe what they see and read. A recent study at Stanford University into how well people evaluate stories on social media came up with this conclusion; “Overall, young people’s ability to reason about the information on the Internet can be summed up in
-
But smaller players are being forced to consolidate to cut costs or close their doors. Consolidation and closures have led to rising unemployment at a time when 2 million young people enter the national workforce each year.⏎“Business is only getting better for the
-
a single brand, blu eCigs, which is owned by the tobacco company Lorillard. The study is the first to extensively analyze trends in youth and young adult exposure to e-cigarette TV ads.⏎“If the current trends continue, awareness
-
40 percent of college-educated workers make less than the top 25 percent of high school graduates.⏎This alternative reading of the figure helps to explain why many young people have not responded so enthusiastically to the “market signals” about college. On
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 most important 500-550 550-600 550-650 550-700 550-750 550-
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
can build apps for different platforms like iOS, Android, or cross-platform frameworks. Proper coding practices ensure that apps are efficient, secure, and scalable, delivering a smooth user experience. With the ever-growing demand for innovative apps, proficient coding skills have
-
).⏎Extension modules play an active role in all these aspects of the Apache server's⏎Figure 3-1. The HTTP transaction consists⏎of a URI request from the browser to the server, followed by a document response⏎from the server to
-
and to transform it into plain-vanilla HTML, XHTML, or other formats (PDF, plain text, rich text files, and so forth) for delivery to Web clients. This helps sidestep the kinds of horrible compatibility issues that dynamic
-
Or, a fluffly feline sits next to a glass of water, representing "tengo sed," or "I am thirsty."⏎The app also challenges its students to answer increasingly difficult typing, match-the-phrase-to-the-photo, listening
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
- L22:2642 strongly excites
- L19:1876 excites
- L22:1721 excites
- L23:1406 suppresses
- L17:2528 suppresses
- L20:2161 suppresses
- L23:98 excites
- L20:1436 suppresses
- L21:1864 excites
- L17:3062 suppresses
- L17:3003 excites
- L23:436 excites
- L19:2840 excites
- L17:1518 suppresses
- L22:712 suppresses
- L16:2041 excites
- L19:2845 suppresses
- L22:1636 suppresses
- L23:1493 excites
- L22:94 suppresses
- L19:2996 suppresses
- L22:496 suppresses
- L19:1970 excites
- L17:1847 excites
Feeds later layers
- L26:657 strongly excites the word 'people' coming next
- L27:2339 excites
- L25:1560 excites
- L27:1515 weakly excites
- L27:1884 weakly excites
- L25:1291 weakly excites
- L27:3069 weakly excites
- L26:552 weakly excites
- L27:1339 weakly suppresses
- L25:707 weakly excites
- L25:2707 weakly suppresses
- L25:1547 weakly excites
- L26:1201 weakly suppresses
- L27:1257 weakly excites
- L25:173 weakly suppresses
- L25:1550 weakly suppresses
- L26:663 weakly suppresses
- L25:1177 weakly suppresses
- L25:2952 weakly excites
- L27:2516 weakly excites
- L26:2543 weakly excites
- L27:1411 weakly excites
- L27:3012 weakly suppresses
- L26:1127 weakly 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
- people
- person
- people
- People
- PEOPLE
- persons
- person
- People
- _people
- (person
Suppresses
- eken
- 候选
- göre
- player
- roker
- Player
- .float
- ecute
- .components
- ajaran