the 're' in self-reliant and similar
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. 26 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 're' piece in words like self-reliant, self-renew, self-repair and recovery, where a hyphen or a word part comes before it. Its opposite side fires on commas in lists just before a final 'and'. Turning it up made the text fall into repeated 'and' and 'or' words.
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, and hydrangea. Rejuvenation pruning is best done in late winter or early spring.<|endoftext|>Neural stem cells (NSCs) can self-renew and differentiate into neurons and glia. Transplanted NS
-
be noted that during the process of clearing the body of toxins, the liver is damaged by exposure to harmful substances. The liver therefore possesses the amazing ability to self-repair and regenerate of damaged tissues.⏎So as one can plainly see, proper liver
-
treating women and children.⏎Women can accomplish good work and I have observed they are always above men in their class. This is accounted for in that she is self-reliant. Women know they will meet opposition, and prepare for it; they
-
to environmentally relevant behaviors. This observation may inspire researchers to⏎more directly test predictions arising from current theoretical models,⏎gather data on actual behavior rather than relying on self-reports,
-
, they’re likely to be fully prepared for snack time but missing important papers or sports equipment. What’s the right thing to do?⏎Your desire to raise self-reliant children is fantastic. But there’s no doubt that passing the baton
-
is typically warmed simply by the heat of the students inside.⏎Ventilation is sized carefully at the correct low volume exclusively to provide appropriate fresh air. The central heat-recovery heat exchanger that makes the once-through air system efficient is the largest piece
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: @@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
to get things done and to keep the United States functioning properly.⏎The Three Branches⏎The three branches of the U.S. government include the executive, legislative, and judicial branches. Each one has its own set of abilities. The legislative branch
-
, men were believed to grow beards for different reasons than men do today. The three main reasons prehistoric men had beards were for warmth, intimidation, and to defend their face from hits. In Egyptian times, men would dye their
-
are created and make sure they comply with the Constitution. It’s easy to see how these groups must work to balance each other as they attempt to create, enforce, and interpret the laws
-
aid in keeping racism alive. Researches strongly concur that minorities are overrepresented in all 3 components of the Criminal Justice System; the police, the courts, and corrections. However, researchers are divided as to why this disproportional representation
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
- L25:6 strongly suppresses
- L19:1245 strongly suppresses comma opening a list of parallel items
- L19:290 strongly excites
- L21:739 strongly suppresses 'both X and Y' pairing words
- L25:1270 strongly suppresses
- L20:2746 strongly suppresses
- L20:852 strongly excites
- L24:205 strongly suppresses
- L23:2137 strongly suppresses
- L18:1856 strongly excites
- L19:2416 strongly suppresses
- L21:2699 strongly suppresses
- L21:64 strongly suppresses
- L17:2529 strongly suppresses
- L20:2812 strongly excites
- L20:1009 strongly excites
- L20:560 strongly excites
- L19:628 strongly suppresses
- L19:1362 strongly excites
- L20:523 strongly excites commas before an elaboration
- L19:3055 strongly excites
- L18:1327 strongly excites
- L19:1092 strongly excites
- L20:2864 strongly suppresses
Feeds later layers
- L27:289 strongly excites
- L27:376 excites
- L27:265 excites
- L27:1902 suppresses
- L27:1698 excites
- L27:94 suppresses
- L27:1937 excites
- L27:258 weakly suppresses
- L27:2212 weakly excites
- L27:191 weakly excites
- L27:52 weakly suppresses
- L27:338 weakly excites
- L27:498 weakly excites
- L27:700 weakly suppresses
- L27:542 weakly excites
- L27:59 weakly suppresses full stops in running text
- L27:1290 weakly suppresses
- L27:1443 weakly suppresses
- L27:690 weakly excites
- L27:2349 weakly suppresses
- L27:425 weakly suppresses
- L27:467 weakly excites
- L27:568 weakly excites
- L27:255 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
- and
- or
- 和
- and
- 或
- And
- And
- và
- 和技术
- -and
Suppresses
- ')">
- @return
- aille
- entirety
- %</
- emsp
- ')</
- nbsp
- &apos
- pomoc