'of' in journal 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. 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 fires on the word "of" inside the names of academic journals, such as "Journal of", "International Journal of" and "American Journal of". It is pushed the other way by minus and plus signs in maths problems. Turning it up did not produce a clear theme in the written text.
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.
-
E., Brown, S., & Ekman, S.L. (2002). Caregiver singing and background music in dementia care. Western Journal of Nursing Research, 24(2), 195-21
-
Dispersal of Aedes Aegypti in an Urban Area After Blood Feeding as Demonstrated by Rubidium-Marked Eggs," American Journal of Tropical Medicine Hygiene⏎52 (2):177-17
-
Marini, L. (2021). Contrasting response of native and non‐native plants to disturbance and herbivory in mountain environments. Journal of Biogeography, 48(7), 1594-
-
. (2012). Social connections and the persuasiveness of viral campaigns in social network sites: Persuasive intent as the underlying mechanism. Journal of Marketing, 18(1), 39-53. http
-
sugars sometimes.⏎3. Green tea.⏎The World Journal of Gastroenterology did research and proved that green tea destroys some bad fats,
-
10.001⏎BERISHA-NAMANI, M. (2013). Information technology, internet, and marketing. International Journal of<|endoftext|>When Jesus lived in His fleshly body, He was alive to the power
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 had a new customer who was a full-time employee of a client who was a major client of a third-time customer who was a client of a customer (a new customer herself) who was a complete customer of a customer (a new
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
1 where n is the number of bits that represent the pixel. Thus, an 8-bit pixel would have a maximum value of (2^8) - 1 = 255. Let the maximum value for pixels in image
-
are perpendicular.⏎Let's see how to identify parallel and perpendicular lines.⏎Let's take some examples to understand this.⏎Considering two equations:⏎y = 2x +3 and y = 2x+5⏎On comparing y = 2
-
that,⏎Lines with the same slope are parallel.⏎Hence, the given lines are parallel.⏎Let's consider another set of equations.⏎y = 2x +3 and 4y = –2x+10⏎On comparing y
-
If we want to iterate an array in reverse order, we need to create a sequence for the for loop using range with:⏎- start as length of the array - 1⏎- stop as -<|endoftext|>12 April 1953
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
- L8:1585 suppresses
- L14:446 weakly excites
- L15:1751 weakly excites
- L9:7 weakly excites
- L13:132 weakly excites "of" after a noun
- L7:117 weakly suppresses
- L0:1620 weakly excites
- L2:2546 weakly suppresses
- L15:310 weakly suppresses
- L11:2882 weakly suppresses
- L14:3030 weakly excites
- L11:936 weakly suppresses
- L8:18 weakly suppresses
- L3:2928 weakly excites
- L14:832 weakly excites modifier words before their head word
- L12:470 weakly suppresses
- L8:1032 weakly excites
- L1:2346 weakly suppresses
- L10:33 weakly suppresses
- L10:2522 weakly suppresses
- L9:41 weakly excites
- L1:1591 weakly suppresses
- L14:1634 weakly excites
- L15:1869 weakly excites
Feeds later layers
- L19:1923 weakly suppresses
- L21:1697 weakly suppresses
- L26:566 weakly excites
- L23:2179 weakly suppresses
- L19:885 weakly excites
- L24:2894 weakly excites
- L20:718 weakly suppresses
- L23:1502 weakly suppresses
- L22:1942 weakly suppresses
- L20:2986 weakly suppresses
- L19:1808 weakly suppresses
- L21:1257 weakly suppresses
- L19:1872 weakly suppresses
- L19:190 weakly suppresses
- L20:1783 weakly suppresses
- L19:787 weakly excites
- L19:301 weakly excites
- L19:1028 weakly excites
- L19:1472 weakly suppresses
- L25:2821 weakly suppresses
- L20:1164 weakly suppresses
- L26:232 weakly suppresses
- L20:1365 weakly suppresses
- L21:440 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
- istributed
- ums
- atter
- (form
- asz
- ubits
- STIT
- distributed
- usan
- enic
Suppresses
- own
- 完整性
- OWN
- 完好
- 专享
- 主动性
- 完整
- hecy
- 占有率
- 发现自己