joins inside names and decimal digits
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. 27 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 at points inside a name or term where a hyphen, bracket or 'of' links two parts, like the bracket in 'Wired Equivalent Privacy (WEP)'. It also goes strongly negative on digits inside decimal numbers. The steered outputs were mixed and did not show one 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.
-
developed by the Wi-Fi Alliance to secure wireless computer networks. The Alliance defined these in response to serious weaknesses researchers had found in the previous system, Wired Equivalent Privacy (WEP).⏎WPA (sometimes referred to as the draft IEEE
-
learn how to use different art mediums and offer your child the opportunity for self-expression. Activities include scribbling with crayons, chalking, clay or Play-doh, easel painting, and watercolor painting.⏎Side Note: Gluing
-
series so PLEASE, PLEASE leave a comment if you LOVE your math series! I'd love to check it out!⏎I like the tricks (9s trick, Doubles, Doubles Plus 1) that Saxon uses to teach facts,
-
a rhyme⏎to rock a rhyme that’s right on time⏎It’s tricky… it’s tricky (tricky) tricky (tricky)”⏎–Run-D.M.C., “It’s Tricky”⏎Long-term planning is hard, especially
-
AYCAT.⏎As part of The World As It Could Be Tree Dedication Ceremony, students, teachers, and other collaborators on the project reflect on the Universal Declaration of Human Rights. Balboa High School Tree Ceremony from BAYCAT.<|endoftext|>Several
-
have been missing. It helps you become a true citizen of the world.⏎History of Global Diversity Awareness Month⏎After the Second World War, The Universal Declaration of Human Rights was adopted by the United Nations General Assembly at the Palais de Cha
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: how can we have strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy strategy
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
= vsin(theta) Immediately after take off : vx = 2cos25 = 1.8126 vy = 2sin25 = 0.845
-
estimated as n log102. Thus, the probability is about cxn, where c = (10/9)2 ≈ 1.2 and x = (9/10)log102 ≈
-
is intended.⏎Btu (British Thermal Unit) — The amount of heat required to raise the temperature of one pound of water one degree Fahrenheit; equal to 252 calories.⏎Charge Controller — A component of a phot
-
the satellite view is shown instead.⏎Latitude: 52.5162 / 52°30'58"N⏎Longitude: -3.0604 / 3°3'37"W⏎OS
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
- L26:1564 strongly suppresses
- L19:753 excites
- L8:10 suppresses
- L8:14 suppresses
- L8:16 excites
- L20:312 excites
- L8:7 suppresses
- L18:569 suppresses
- L24:226 excites
- L6:13 suppresses
- L25:11 excites
- L22:201 suppresses
- L8:4 suppresses
- L1:2677 excites
- L8:13 excites
- L6:8 excites
- L8:3 suppresses
- L7:62 suppresses
- L6:18 suppresses
- L5:216 excites
- L18:576 excites
- L6:12 suppresses
- L22:225 excites
- L20:52 excites
Feeds later layers
None: this is the last layer.
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
- urers
- ellers
- Malays
- USERNAME
- ocache
- MMC
- TTC
- .PERMISSION
- antal
- кли
Suppresses
- โอกา
- đự
- וסיף
- nawet
- ในฐานะ
- 旗舰店
- ,...⏎⏎
- ım
- +,