opening parenthesis before an acronym
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 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 opening bracket that comes right before a short acronym, such as the '(' in 'Resting Metabolic Rate (RMR)'. Turned down, the model produced garbled repeated words. Turned up, it wrote text that did not clearly match.
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.
-
during the oxidation of food and carbon dioxide (CO ) is produced. 2⏎), in proportion to the heat production.⏎Measuring Resting Metabolic Rate (RMR) is fast and simple. It includes a short, non-invasive
-
made in laboratories around the world. And they are among the goals of more than 100 horizon-stretching research projects housed in the Advanced Measurement Laboratory (AML) Complex at the National Institute of Standards and Technology (NIST).⏎Completed
-
of progressive science education centres within the Africa-Mediterranean-European (AME) area.<|endoftext|>THE NEW FOUNDE LAND⏎Volume 18
-
, and gave glycerin-supplemented poultry feed to broilers that were 7 to 45 days old.⏎Dozier evaluated the apparent metabolizable energy (AME) intake—a standard measure of energy—for the study group and for a control
-
vibrations (<100 Hz) are naturally generated by skeletal muscles during voluntary contractions. Recording of these vibrations at the muscle surface are called surface mechanomyograms (S-MMGs). In this study, S-MMGs were recorded over a
-
. https://www.academia.edu/39903340/Displaying_an_Icon_The_Mosaic_Icon_of_Saint_Demetrios_at_Sassoferrato_and_its_Frame
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: remmorum remmorum rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
After the Romans London was conquered by the Angles and the Saxons who brought their language with them (Old English)⏎In 1066 William the Conqueror conquered London with his Norman (though French speaking) army. Not only
-
quest, documented history starts around the end of the 13 th century, when the hall was passed to the Earl of Chester, a nephew of William the Conqueror.⏎Since the year
-
"?⏎Yes, curling has been compared to chess by fans and detractors alike. Fans emphasize the strategy involved. Detractors claim that both sports
-
hands, (Poland having been invaded by Germany), he has hidden it.<|endoftext|>BURKE TEACHERS STUDY, PREPARE FOR NEW CURRICULUM⏎Burke County schools are preparing for a curriculum change coming in August as
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
- L25:1061 strongly suppresses
- L22:1810 excites
- L25:925 excites
- L25:2560 suppresses
- L24:1201 excites
- L25:237 excites
- L24:2795 excites
- L21:133 suppresses
- L23:1793 weakly excites
- L19:334 weakly excites
- L19:1581 weakly suppresses
- L23:1334 weakly suppresses
- L24:1038 weakly suppresses
- L5:1397 weakly suppresses
- L0:2560 weakly excites
- L12:1357 weakly excites
- L10:2114 weakly suppresses
- L6:940 weakly excites
- L12:3001 weakly excites
- L17:1059 weakly suppresses
- L4:2018 weakly excites
- L25:2378 weakly suppresses
- L23:188 weakly excites
- L7:1368 weakly excites
Feeds later layers
- L27:422 suppresses
- L27:464 suppresses
- L27:1012 suppresses
- L27:941 suppresses
- L27:211 excites
- L27:790 weakly suppresses
- L27:3027 weakly suppresses
- L27:2995 weakly suppresses
- L27:2288 weakly suppresses
- L27:2608 weakly suppresses
- L27:1515 weakly suppresses
- L27:720 weakly suppresses
- L27:1933 weakly suppresses
- L27:89 weakly excites
- L27:87 weakly suppresses
- L27:2771 weakly suppresses
- L27:2796 weakly suppresses
- L27:2178 weakly suppresses
- L27:1359 weakly suppresses
- L27:590 weakly excites
- L27:178 weakly excites
- L27:2596 weakly suppresses
- L27:1426 weakly suppresses
- L27:991 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
- <![
- <!--[
- <!--[
- ưỡ
- ▩
- 只会
- hasNext
- canActivate
- 뜨
- viewDidLoad
Suppresses
- em
- Emm
- rem
- EM
- rem
- Em
- EM
- Em
- um
- (em