start of a new document
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 on the separator the training text uses between web pages, specifically on the first word after that separator. It also turns slightly negative on the word 'fructose'. Steering made the output repetitive without a clear link to the trigger.
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.
-
your own microclimate," architect and Natural Home columnist Carol Venolia says. "Working with nature is simple, but we lost touch with that ability when we became<|endoftext|>THE NEED FOR CULTURAL REVOLUTION⏎Prof. Jose Ma. Sison
-
smartest thing you ever did.Powered by Sidelines<|endoftext|>There is a wealth of scientific literature available on the treatment of adults diagnosed with schizophrenia
-
<|endoftext|>House in Space, from a NASA Remix Challenge. Credit: Cookieater20
-
<|endoftext|>10 Tips for Parents of Prospective College Students⏎1. Choosing a Career
-
<|endoftext|>Active transportation refers to non-motorized transportation modes, such as bicycling and
-
<|endoftext|>Public holidays in Pakistan are different than bank holidays in the country. While all of
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 current ) as function 1 1. What system What What What topics are: the question: questions: What titles. The questions you will questions. What titles: questions.
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
fructose-1-6⏎diphosphate deficiency and in fructose-1-phosphate aldolase deficiency⏎(hereditary fructose intolerance). Fructose-1-6 diphosphate deficiency⏎produces accumulation of lactic and
-
diet. Continuous vomiting leads to coma. Carbohydrate abnormalities are⏎diagnosed by finding reduced enzyme activity in the liver. Treatment consists⏎of eliminating fructose from the diet and correcting hypoglycemia.⏎syndrome is produced
-
and potatoes are classified as complex carbohydrates. However, the body converts starches to blood sugar as rapidly as it processes pure glucose. On the other hand, fructose (fruit sugar) is branded as a simple carbohydrate, yet has a minimal effect
-
, so it must first sever the chemical link connecting the two sugars. The enzyme sucrase in your small intestine assists with the breakdown of sucrose into fructose and glucose. This allows your body to absorb them, transport them to the liver
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
- L1:25 excites
- L1:2 excites
- L3:24 excites
- L1:7 excites
- L3:1256 excites
- L1:57 excites
- L1:109 excites
- L1:49 excites
- L3:1 suppresses
- L1:2554 excites
- L5:931 suppresses
- L4:11 excites
- L1:1227 suppresses
- L3:8 excites
- L1:42 suppresses
- L16:769 excites
- L1:2618 excites
- L3:504 excites
- L2:52 excites
- L1:751 suppresses
- L2:2128 excites
- L1:2012 excites
- L2:47 excites
- L5:574 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
- -------------</
- 第二个
- 什么
- ROUND
- THON
- омн
- жен
- PLIC
- *</
- ;'⏎
Suppresses
- ettes
- 商报
- 鳅
- like
- nowadays
- checker
- blick
- lette
- iever
- plus