opening bracket after a unit conversion
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. 20 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 reacts to the opening bracket that comes right after a measurement, where the writer gives the same amount in another unit, like '5 in. (12.7 cm)'. Turning it up didn't make the replies about measurements, so the link shows only in the text it reads.
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.
-
The Central Business District as seen from the top of the Carlton Centre⏎|Municipality||City of Johannesburg|⏎|• Total||1.09 km2 (0.42 sq mi)|⏎|• Density||13,0
-
than 30 minutes after shutdown⏎- Use clean, sealed bottles and flush<|endoftext|>The capture envelope for this system is approximated by a 5-in. (12.7-cm) cube.⏎A prototype system for docking a miniature
-
docking port. The capture envelope for this system is approximated by a 5-in. (12.7-cm) cube centered on the front of the docking-port
-
bps. The net result of these simultaneous forces is a 25% increase in monthly transit fees, from $16,000 per month to $20,000 per month. The percentages vary
-
Visayan warty pigs seasonally grow large manes that start at the top of their head and travel partially down their back and grow up to 9 inches (23 centimeters) long! The mama of the group is Kurit (
-
which caused the deadliest tsunami in modern history – had such an effect.⏎Gross calculated that the Chilean quake shifted the Earth's figure axis by 3 inches (8 cm or 27 milliarcseconds), and shortened the length of
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: the [map] is a [map] in the [map [map [map [map [map [map [map [map [map [map [map [map [map [map [map [map [map [map [
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
Germs and other things that cause foodborne illness usually can be killed when heated. Heating food to an internal temperature above 160 degrees Fahrenheit or 78 degrees Celsius, even for a few seconds, is enough to kill most bacteria
-
3.0°F (22.8°C). The warmest month, on average, is November with an average temperature of 81.0°F (27.2°C). The coolest month on average is June, with an
-
inches of rainfall in 2016, the temperature in this arid region varies from freezing cold in winter to over 100º F (38º C) in summer.⏎It was also home to some of the oldest
-
6 25.........⏎Basics: Divide integers to get integer outputs in all cases⏎Determine value using Look-up table method⏎Create Volcano (or Atoll) martix. It is an extension of Bullseye matrix problem
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
- L3:1911 excites
- L12:1332 weakly suppresses
- L10:1950 weakly excites
- L4:2313 weakly suppresses
- L9:1204 weakly suppresses
- L3:1312 weakly suppresses
- L6:2393 weakly suppresses
- L5:1704 weakly excites
- L13:2541 weakly suppresses
- L5:27 weakly excites
- L10:157 weakly suppresses end of document and web boilerplate
- L7:2440 weakly excites
- L17:1506 weakly suppresses
- L3:594 weakly suppresses
- L13:2186 weakly excites
- L1:779 weakly excites
- L5:86 weakly suppresses
- L6:1390 weakly excites
- L9:318 weakly suppresses
- L13:1723 weakly suppresses
- L9:2331 weakly excites
- L12:2617 weakly suppresses
- L17:172 weakly excites
- L17:1582 weakly suppresses
Feeds later layers
- L22:2388 suppresses
- L22:2840 suppresses 'as' in 'as low as' and 'as far as'
- L23:2845 weakly suppresses
- L21:2323 weakly suppresses
- L21:1848 weakly suppresses
- L21:2151 weakly excites
- L21:2637 weakly excites
- L23:1035 weakly suppresses
- L22:1671 weakly excites
- L25:1747 weakly suppresses
- L27:2916 weakly suppresses
- L21:995 weakly suppresses
- L24:2631 weakly suppresses
- L22:2747 weakly suppresses
- L27:2006 weakly suppresses
- L22:2360 weakly excites
- L27:2885 weakly suppresses
- L25:2292 weakly excites
- L27:1344 weakly suppresses
- L21:520 weakly excites
- L23:861 weakly excites
- L25:1353 weakly excites "in" before "good" condition
- L26:2143 weakly excites
- L22:2678 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
- ائية
- ски
- emia
- presence
- 疡
- presence
- 祓
- /gin
- ском
- {name
Suppresses
- map
- maps
- map
- .map
- -map
- (map
- →map
- Map
- apply
- maps