number sequences and patterns
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. 17 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them 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 blank spot in a number pattern, the place where the next number in a sequence should go, such as in '1, 1, 2, 3, 5, [8]' or '4 = 9 minus 5'. Turning it up made the text drift into random digits and repeated number strings.
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.
-
a matrix⏎Find nth maximum⏎Matlab Basics II - Create a vector with a repeated entry⏎Create sequnce 1 4 9 16 25.........⏎Basics: Divide integers to get
-
3, 4, 5], it will print it in reverse order i.e. from the last element to the first element:⏎5 4 3 2 1⏎We will use a loop to do that.⏎The values are
-
awareness is when a child can hear and make rhymes. Corps members and children read poems together. They listen for words that rhyme, like “no” and “so.”⏎At the core of Jumpstart session plans is a set of 2
-
this rule (as in soliloquies).⏎Add -s to most nouns ending in -f.⏎Exceptions: In some cases, change the -f to -v and add -es⏎In compound words, make the main word plural.
-
’t recognize the logic of that switch, review how fact families work by using smaller numbers: 5 = 9 – 4 means also that 4 = 9 – 5.⏎Made = 650⏎Sold = ?⏎Left
-
We are also learning about increasing and decreasing patterns and figuring out what the pattern rule is. For example, a decreasing number pattern would be 850, 845, 840, 835, 8
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 want to make a series of 1000000000000000000000000000000000000000
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
, dairy, and eggs, vegans consume a cholesterol-free diet. The type of protein in a vegetarian diet may be another important advantage. Many a Healthy Weight .10 studies show that replacing animal protein with plant protein lowers blood cholesterol levels—even
-
Limits 1. sin lim 1 x x x = 2. 1 cos lim x x x -= 3. ( ) 1 lim 1 x x x e + = Strategy for Finding Limits: 1. Try direct
-
fills and widens cracks and fractures as it rises. The end result is a vertical wall, or dyke,<|endoftext|>-Why is balance important in an ecosystem?⏎-What is the relationship between living and non-living things in an ecosystem?⏎
-
of the University's work as a Spoke, is one of the leading UK university groups in its field, spanning four<|endoftext|>- What are some common folk tales?⏎- Are folktales true stories?⏎- What are the 5 types of
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
- L15:1617 excites
- L15:1448 suppresses
- L16:2131 excites
- L0:2922 weakly suppresses
- L12:1144 weakly excites
- L9:246 weakly suppresses
- L15:868 weakly excites
- L8:1049 weakly suppresses
- L13:2600 weakly suppresses
- L1:3054 weakly excites
- L15:1533 weakly suppresses
- L11:2366 weakly excites
- L7:133 weakly suppresses line break inside a sentence
- L5:2062 weakly excites
- L12:2045 weakly suppresses
- L6:1447 weakly suppresses
- L12:385 weakly suppresses
- L0:1490 weakly excites
- L9:2703 weakly suppresses
- L14:1432 weakly excites
- L11:947 weakly suppresses
- L4:2693 weakly suppresses
- L8:2919 weakly excites
- L9:1752 weakly suppresses
Feeds later layers
- L18:1373 suppresses
- L25:1534 weakly excites
- L24:1650 weakly excites
- L18:2153 weakly suppresses
- L18:352 weakly excites
- L21:1703 weakly suppresses
- L27:126 weakly excites
- L20:1853 weakly excites
- L19:2590 weakly excites
- L19:614 weakly excites
- L25:1790 weakly excites the digit 1 starting a statistic
- L22:508 weakly excites
- L22:1300 weakly suppresses
- L22:1520 weakly excites
- L25:963 weakly excites
- L24:2959 weakly suppresses
- L22:105 weakly suppresses
- L22:86 weakly excites
- L21:2953 weakly suppresses
- L18:2333 weakly excites
- L23:3069 weakly excites
- L22:386 weakly suppresses
- L23:802 weakly excites
- L25:484 weakly suppresses
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
- 동
- JKLMNOP
- imiters
- Handles
- ído
- upt
- weets
- ucid
- EFR
- tória
Suppresses
- чи
- chỉnh
- pouch
- rút
- nhi
- gun
- verse
- uri
- 美术馆
- uom