'and' joining two nouns in a pair
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 lights up on the word 'and' when it links two people, places, or ideas in a pair, such as 'X and Y'. Turning it up made the text wander into loose, repetitive phrasing.
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.
-
amendments granting these regions formal recognition and autonomy.⏎In 1948 Belgium became a co-signatory to the Benelux Customs Union along with the Netherlands and Luxembourg. This later became the Benelux Economic Union in 195
-
opper showed early interest in art, particularly drawing, and went on to study illustration and painting. With their emphasis on truthful, contemporary subjects, his teachers Robert Henri and Kenneth Hayes Miller at the New York School of Art were vital to Hopper's
-
Southern states-Virginia, North Carolina, Arkansas, and Tennessee-joined the Confederacy once the war started. Secessionist governments were established in Missouri and Kentucky, two border states that officially remained in the Union, while the western counties
-
-Nicod within France's National Center for Scientific Research (CNRS) and a Global Distinguished Professor at New York University.⏎The combined team of linguists and primatologists analyzed alarm calls of Campbell's monkeys on two sites: the Tai
-
and can be used to add visual interest to an outfit. When choosing patterns, it is important to consider their scale and how they will work with other elements of your outfit. You can also mix and
-
an archaeologist with a strong background in anthropology and museums, passionate about outreach and education. Her research focuses on death and burial,
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: what I want to do with this as a a final, only final effort to build it up, and then then then then then and then and then and then and then and then and then and then and then and then and then and then
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
that planets orbit at least 5 to 10 percent of all stars. But how many of these planets are Earth-size, and possibly Earth-like?<|endoftext|>& Tornado Alley⏎Jan Mikusinski developed an operational calculus which is relevant for
-
. It is a free software. We also strongly recommend that you sign up for an AdobeID at the Adobe website. For more details please see FAQ 1&2. To view this ebook
-
. It is a free software. We also strongly recommend that you sign up for an AdobeID at the Adobe website. For more details please see FAQ 1&2. To view this ebook on an iPhone, iPad or Android mobile device you
-
across the United States. Others are limited to specific regions of the country. Thankfully, however, most ticks are fairly easy to identify.⏎Genus & Species⏎& Conditions Caused⏎|American Dog Tick⏎||Dark brown with creamy-gray markings
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
- L13:131 excites 'and' joining parallel phrases
- L4:57 suppresses
- L9:92 suppresses "and" or "or" joining two items
- L15:114 excites
- L1:1250 suppresses
- L0:2776 suppresses
- L15:82 suppresses
- L3:447 excites
- L5:548 excites
- L15:115 excites
- L10:74 excites
- L15:78 weakly excites
- L16:177 weakly suppresses zero digit inside dates and codes
- L17:1950 weakly suppresses
- L16:135 weakly suppresses
- L16:208 weakly suppresses
- L16:503 weakly excites
- L16:376 weakly suppresses 'or' joining alternatives
- L15:89 weakly excites
- L15:80 weakly suppresses
- L15:122 weakly suppresses
- L16:282 weakly excites
- L8:7 weakly excites
- L16:34 weakly excites
Feeds later layers
- L22:134 suppresses
- L20:885 suppresses 'and' joining paired words
- L20:2504 excites
- L25:1694 suppresses 'and' joining paired names or fields
- L19:2080 suppresses ampersand joining two items
- L23:270 excites
- L20:1536 weakly suppresses the word 'well' in phrases
- L22:2880 weakly suppresses
- L19:2565 weakly excites
- L20:2371 weakly excites
- L22:198 weakly suppresses
- L23:821 weakly excites
- L19:334 weakly excites
- L21:1246 weakly excites
- L19:860 weakly excites
- L25:2143 weakly suppresses
- L21:921 weakly excites
- L20:533 weakly excites
- L27:2695 weakly suppresses
- L20:3043 weakly suppresses
- L21:1765 weakly suppresses
- L21:357 weakly excites
- L21:1546 weakly suppresses
- L23:2057 weakly suppresses ampersand inside names and abbreviations
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
- че
- 客户提供
- 철
- ading
- ราม
- 遷
- ない
- 一层
- 出入
- 买单
Suppresses
- amp
- lt
- emsp
- raquo
- nbsp
- quot
- nds
- gt
- uum
- vice