the 'S' in 'U.S.'
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. 0 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 weakly on the 'S' in the abbreviation 'U.S.', as in 'U.S. Food and Drug Administration'. The effect is small, and turning it up gave off-topic text, so the steering does not show much.
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.
-
Often, one might completely overlook the positives<|endoftext|>If you take your ears for granted, listen up: hearing loss is the third most common health problem in the U.S. It's also on the rise. Compared to a 1971
-
article helpful?<|endoftext|>Given that this is the first U.S. presidential election since apps have made their way onto most electronic devices, you might
-
Muslims form a small but important segment<|endoftext|>Gardenia Care & White Flies⏎Although gardenias (Gardenia spp.) are recommended for growth in U.S. Department of Agriculture plant hardiness zones 8 and above, cool temperatures are
-
see Question 7).⏎- Laetrile is not approved by the U.S. Food and Drug Administration (FDA) (see Question 8).⏎Questions and
-
find the percentage of a whole number.<|endoftext|>1. How does the U.S. manage the environmental legacy and human health impacts of its nuclear weapons?⏎The U.S. nuclear weapons program has had a wide range of detrimental effects to both human health
-
). Thanks to the achievements of public health, we live longer and healthier lives. According to the National Center for Health Statistics, “life expectancy at birth among U.S. residents increased by 62%, from 47.3 years in
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 have been working on a project that is a little of a small project. It's a small, but it has a large, and it is a big, and it has a big, and it is a small, and it is
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
to better accommodate your needs within the guidelines of the (GNU Public License). The commercially licensed version is available for more flexibility and advanced support.⏎So How Does MySQL Work?⏎One or more devices (clients) can connect to a server through a specific
-
Associate Professor of Philosophy at the University of Texas at Arlington.⏎—Lawrence Zbikowski, Department of Music, University of Chicago⏎—Anthony Cascardi, University of California, Berkeley⏎—Jesse J. Prinz,
-
his/her partner by matching the Valentine halves.⏎Similar activities:Valentine's Day Heart Envelope, This Much Valentine's Day Card, Giving Hand Valentine's Day Card, Valentine's Day Photo Card, Valentine's Day Hug Me Card, Make
-
the lender will be assessing the borrower's ability to repay the loan based on a worst-case scenario interest rate. The calculation for individuals with irregular income may be different from those
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
None: this is the first layer.
Feeds later layers
- L1:41 strongly excites
- L2:4 strongly excites
- L1:0 excites
- L3:36 excites
- L2:878 suppresses
- L3:2 excites
- L1:7 excites
- L1:22 excites
- L3:8 excites
- L18:335 weakly excites
- L1:2 weakly excites
- L4:503 weakly suppresses
- L3:440 weakly suppresses
- L2:52 weakly excites
- L26:1170 weakly suppresses
- L1:84 weakly suppresses
- L17:83 weakly suppresses
- L19:276 weakly suppresses
- L23:1314 weakly suppresses
- L2:47 weakly excites
- L6:283 weakly excites
- L21:85 weakly suppresses
- L1:842 weakly suppresses
- L24:2421 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
- ds
- atch
- 噪
- тельных
- ��
- aves
- ilot
- 噪声
- ly
- 昼
Suppresses
- curl
- nze
- urn
- rigged
- vul
- HttpRequest
- ооруж
- ולי
- KEEP
- Sext