unclear: url fragments and math variables
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. 10 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 mix of things: the 'text' part of a web link, the letter t in a math formula, and some colons and dots. The examples don't share one clear idea. Steering produced text with no obvious connection to the pattern.
Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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.
-
une Function, 2017, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5707683/#:~:text=It%20is%20a%20potent%
-
What bumper means?⏎Definition of bumper (Entry 3 of 3) 1 : a device for absorbing shock or preventing damage (as in collision) specifically : a bar at<|endoftext|>Like other Western states, New Mexico is home to wide open
-
Power of a sinusoidal waveform⏎Consider an electrical AC sinusoidal signal which is characterized by its voltage V(t)=Vmax×sin(ωt+ΦV) and current I(t)=Imax×sin(ωt
-
is available for this purpose (e.g., US Census data at https://www.census.gov/) and can be of tremendous help in targeting various ethnic groups and geographic areas with
-
power is then written as⏎Here the average of sin2 ωt over a cycle is 1/2 and that of sin ω t cos
-
-835-5322 (TTY) or at email@example.com.<|endoftext|>Table of Contents :⏎Top Suggestions Senior Kindergarten Writting Worksheet :⏎Senior Kindergarten Writting Worksheet In this worksheet your students will practice their handwriting
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 am going to write a new function to calculate the area of a circle." So, I need to calculate the area of a circle. I need to find the area of a circle in radians. The area is equal to $2
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
4×3×2×1 is written as 8! and we call it "eight factorial." You probably have a button on your calculator labelled with [x!] that will perform these calculations for you.⏎When you shuffle a pack of cards
-
parent graph of the cotangent function f(x) = cot x,⏎by using the same techniques you use to find the tangent parent graph
-
exam will be done. Tests may include:⏎Treatment focuses on strategies to improve the immune system. If you have HIV, your doctor will most likely prescribe antiretroviral medications to treat this condition. If PML has resulted from the drug
-
whether u-substitution might be appropriate. Your first temptation might to said,⏎- "Hey, maybe we let u equal sine of 5x, and if u is equal to sine of 5x,⏎- we have something that is pretty
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
- L9:1177 suppresses
- L8:1486 excites
- L7:2335 weakly suppresses
- L9:2823 weakly suppresses
- L9:2026 weakly suppresses
- L9:893 weakly suppresses
- L6:847 weakly excites
- L2:2217 weakly suppresses
- L8:1880 weakly excites
- L7:791 weakly excites
- L5:1074 weakly suppresses
- L6:482 weakly suppresses
- L3:2252 weakly excites
- L8:7 weakly excites
- L3:497 weakly excites email addresses at the end of text
- L9:1135 weakly suppresses
- L4:83 weakly excites
- L7:2431 weakly suppresses
- L4:1212 weakly suppresses
- L9:2078 weakly excites
- L1:512 weakly excites
- L9:1236 weakly suppresses
- L6:1370 weakly suppresses
- L7:472 weakly excites
Feeds later layers
- L14:1287 weakly suppresses
- L12:1471 weakly suppresses
- L22:2940 weakly excites
- L18:2163 weakly excites
- L12:2777 weakly excites
- L12:2208 weakly suppresses
- L12:2915 weakly suppresses
- L15:711 weakly suppresses
- L16:698 weakly excites
- L24:2131 weakly excites
- L18:2415 weakly suppresses
- L14:307 weakly suppresses hyphen inside phone numbers
- L13:1723 weakly suppresses
- L20:619 weakly excites
- L11:535 weakly suppresses
- L16:1522 weakly suppresses
- L13:2693 weakly suppresses
- L20:2433 weakly excites
- L12:1905 weakly suppresses
- L20:995 weakly suppresses
- L14:2404 weakly suppresses
- L11:743 weakly suppresses
- L14:1775 weakly suppresses
- L21:1238 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
- omi
- 问他
- 出发
- apsed
- ”⏎⏎
- friend
- gui
- Violation
- ”:
- ”⏎⏎
Suppresses
- 岘
- luent
- 性强
- tte
- 変わり
- ści
- -piece
- waż
- Ink
- ilda