word before a number, range or quantity
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. 22 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 fires on words that are followed by a number or a measured amount, such as 'maintained at' before a pH value or 'the difference between' before a term. Turning it up made the model's writing drift into unrelated topics, with no obvious pattern.
Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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.
-
individual and physical examination must be performed in a respectful and culturally sensitive manner. Privacy is essential and the nurse must be aware of their own ...⏎You have reached your viewing limit for this book.⏎What people are saying - Write a review⏎Respir
-
your computer uses DNS to retrieve the website’s IP address of 2001:41c8:20::19.⏎We Focus On:⏎Behind
-
17. Draw a diagram showing optimum combination of consumer.⏎18. Mention any two characteristics of oligopoly market⏎19. Clarify the difference between average revenue and marginal revenue.⏎20. Calculate the ex-ante aggregate demand,
-
80 - 120 ppm. The pH is the measure of the acidity of the swimming pool water, and the pH level should be maintained at 7.2 - 7.6 ppm. The pH of pool water is
-
It can be watched in many languages and these can be access by clicking on the backwards and forwards buttons (that look like |< or >|) after you have pressed play.⏎Tiny Happy People is a website dedicated
-
as factors in Parkinson's pathophysiology, multiple articles describing nuances of deep brain stimulation, and stories about exceptional athletes living with Parkinson's.⏎Read on for all of November's Parkinson's news!⏎SCIENCE AND RESEARCH Updates⏎- Chu et
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: "rationala ravenasravena ravena ravena ravena ravena ravena ravena ravena ravena ravena ravena ravena ravena raven
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
interdental cleaning such as flossing or using the Waterpik, can get lost in the hustle. The ugly truth is that this simple task gets put on many people’s low priority list and can be the very reason their oral health
-
of credibility. It would not be an exaggeration to say that from ancient times until now, there has not been a period in which this question has not been put to a writer or an artist without regard to his nationality or the motives for his
-
illustration of how one subject is put for another is found in comparing the two seemingly contradictory biblical accounts of the death of
-
all directions. This may help to make some furtive words more readable. Also the spacing, boldness and capital case can be switched to make the letters stand out from their context. Optionally a handy hover-cross is shown, highlighting those letters
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
- L20:959 suppresses
- L21:237 excites
- L21:746 excites
- L18:569 suppresses
- L16:582 suppresses
- L19:2987 suppresses
- L17:152 suppresses mid-word fragments, often 'intric'
- L16:222 suppresses
- L5:50 suppresses
- L11:42 excites
- L21:475 suppresses
- L7:9 excites
- L7:110 excites
- L3:115 excites
- L2:75 excites
- L8:10 suppresses
- L18:396 excites
- L11:300 excites
- L10:1651 suppresses
- L13:1638 suppresses
- L6:13 suppresses
- L8:7 suppresses
- L18:871 suppresses
- L20:1285 excites
Feeds later layers
- L23:254 excites
- L27:160 suppresses
- L24:385 excites
- L25:206 excites
- L27:342 excites
- L26:495 excites
- L23:154 excites
- L26:685 suppresses
- L26:442 suppresses
- L24:34 suppresses
- L23:760 suppresses
- L23:630 excites
- L23:25 excites
- L25:1493 suppresses
- L24:2423 weakly suppresses
- L23:1613 weakly suppresses
- L24:105 weakly excites
- L27:50 weakly excites
- L26:1041 weakly suppresses
- L24:943 weakly suppresses
- L23:385 weakly suppresses
- L27:766 weakly suppresses word endings split mid-word, like 'shower' in 'showered'
- L27:792 weakly suppresses
- L26:210 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
- Professional
- iterations
- setObject
- 梅州
- 轮融资
- Worksheets
- 发展格局
- Permissions
- 经开区
- Psychological
Suppresses
- nels
- zyć
- gado
- gio
- nement
- naire
- ties
- keiten
- thers
- rious