numbers after 'y =' or 'cosine of'
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. 25 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 a specific number that comes right after a word like 'y =' or 'cosine of', as in 'y = 2x' or 'cosine of 5x'. Turning it up made the model's output drift into long, rambling word salad, and it did not produce a clear theme.
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.
-
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 = 2x +3 and y =
-
have our du.⏎- Our du is 5 -- let me circle that and let me do that in that blue color --⏎- is 5 cosine of 5x dx. So we can rewrite this entire expression as --⏎
-
dx--⏎- sorry, I need cosine of 5x dx, just like that. So when you look over here,⏎- you have a cosine of 5x dx,
-
00000000 IP: 10101100.0001000x.xxxxxxxx.xxxxxxxx Each subnet has 2500
-
00 n = 15 NETWORK IP: 10101100.0001000x.xxxxxxxx.xxxxxxxx SUBNET IP: 10101100.0
-
we know 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
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: cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive cognitive
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
middle school students.19<|endoftext|>毋忘在莒(Pinyin: Wu Wang Zai Ju) is a famous Chinese idiom (成語, Pinyin: Cheng Yu). It is sort of like the Latin idiom left in the European
-
legal system, early Confucian scholars emphasized the concepts of morality and ritualism. The term “ritual propriety” (禮, pinyin: lǐ), describes the “proper” social relationships and the set of rituals which regulate them
-
We hold these Truths to be self-evident, that all Men are created equal, that they are endowed by their
-
, Seobeol). In 503, King Jijeung standardized on the characters 新羅(신라), which in Modern Korean is pronounced "Shilla
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
- L24:495 strongly suppresses
- L23:1915 excites
- L23:1848 excites
- L23:2224 excites
- L22:918 suppresses
- L23:1668 suppresses
- L22:411 suppresses
- L24:2291 suppresses
- L24:1119 suppresses
- L24:2406 suppresses
- L20:2900 excites
- L23:387 suppresses
- L23:1128 excites
- L21:972 excites
- L22:859 suppresses
- L10:1510 excites
- L23:2862 excites
- L23:1764 excites
- L22:1121 suppresses
- L23:2997 weakly suppresses
- L21:2318 weakly suppresses
- L3:2159 weakly excites
- L20:2955 weakly excites
- L20:1450 weakly suppresses
Feeds later layers
- L27:560 strongly suppresses
- L27:1128 strongly suppresses
- L27:589 strongly suppresses
- L27:534 strongly suppresses
- L26:1213 suppresses
- L26:957 suppresses
- L27:768 weakly suppresses
- L26:1038 weakly excites
- L27:1147 weakly suppresses
- L26:293 weakly excites
- L26:1278 weakly suppresses
- L27:1441 weakly suppresses
- L26:162 weakly excites
- L26:2495 weakly suppresses
- L27:566 weakly suppresses
- L27:2795 weakly suppresses
- L27:1080 weakly excites
- L27:1517 weakly suppresses
- L26:2784 weakly suppresses
- L26:1327 weakly excites opening parenthesis before an acronym
- L27:2858 weakly suppresses
- L27:2031 weakly suppresses
- L27:89 weakly excites
- L27:2199 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
- @register
- maids
- ugin
- semblies
- Posting
- erner
- 渤
- 楣
- esign
- ificant
Suppresses
- X
- X
- x
- >X
- x
- ,X
- .X
- _X
- -X
- →X