physical measurements and dimensions
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. 22 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 words and numbers that describe the size of something, like '4 inches on each side', 'a diameter of 5,150 km', or '10 ft in diameter'. Turned down, the text got stuck repeating words like 'area' and 'volume', and turned up it wrote about numeric problems.
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.
-
reach can trap nestlings inside a house even when they need to leave and enter the outside world.⏎- Interior Floor Space: A square interior floor measuring four inches on each side is the best spacing
-
calculate a different measurement. Ask students what additional information they need to find the total amount of cardboard in the box. When students recognize that they need the lengths of the diagonal sides of the box give them the measurements for those sides (2.2
-
|• Land||96.6 sq mi (250.2 km2)|⏎|• Water||0.7 sq mi (1.7 km2)|⏎|Elevation||1,148 ft (
-
vertical orientation of the sculpture references the old-growth trees that once covered the landscape near the museum. Its tapered shape—roughly 10 ft in diameter at its base and approximately 4 ft in diameter at its top— recalls a ship
-
to calculate and return the volume of a sphere based on the given radius.⏎- Your second subroutine should be called getVolumeOfCuboid(), take three parameters called width, length and height and apply the relevant formula to calculate and return the volume
-
by the thickness of the membrane. At steady state, Fickian diffusion can be approximated by:⏎J = D (CWaterlooEmitter - groundwater) / membrane wall thickness⏎where D is the diffusion coefficient for the membrane material and is
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 area around the volume and surface area of the volume, the volume and surface area volumes, and the volume and surface area volumes, and the volume and area volumes, and the area and volume volumes, and the volume and volume volumes,
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
take one parameter called radius and apply the relevant formula to calculate and return the volume of a sphere based on the given radius.⏎- Your second subroutine should be called getVolumeOfCuboid(), take three parameters called width, length and height and
-
, they calculate the surface area of the prism. This provides students with the opportunity to work with complex shapes to find surface area in a given context.⏎Display the image for all to see throughout the
-
: Revisiting the Box of Chocolates (10 minutes)⏎In this activity students are presented with a prism that was used in a previous lesson to calculate volume. Here
-
out (performed) one after another.⏎We will use the following steps:⏎- Convert all measurement to the same unit (e.g. mm)⏎- Calculate the volume of planet Earth⏎- Calculate the volume
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
- L9:2347 suppresses
- L11:1255 excites
- L20:2483 excites
- L20:815 suppresses
- L20:1203 excites
- L6:1887 suppresses
- L10:825 suppresses 'B' in vitamin B names
- L9:2044 suppresses
- L8:1466 excites
- L7:1001 weakly suppresses
- L6:2569 weakly excites
- L10:1369 weakly excites
- L20:402 weakly excites
- L8:2079 weakly excites
- L12:1533 weakly excites
- L5:783 weakly suppresses
- L10:1640 weakly excites
- L6:779 weakly suppresses
- L15:1586 weakly excites
- L9:628 weakly suppresses
- L17:2890 weakly excites
- L9:1335 weakly excites
- L10:2277 weakly excites
- L20:1079 weakly suppresses
Feeds later layers
- L27:1910 suppresses
- L26:2464 suppresses
- L23:904 suppresses
- L25:1853 suppresses
- L23:2849 suppresses
- L27:1729 suppresses
- L24:621 suppresses
- L27:2698 weakly suppresses
- L24:1235 weakly suppresses
- L26:2349 weakly suppresses
- L26:2384 weakly suppresses
- L24:795 weakly suppresses
- L24:1664 weakly suppresses
- L24:550 weakly suppresses surname split after first letter D
- L25:3061 weakly suppresses
- L27:1805 weakly excites
- L24:3035 weakly suppresses
- L23:2167 weakly suppresses
- L24:1708 weakly excites
- L25:1803 weakly suppresses
- L25:2790 weakly excites
- L23:2263 weakly suppresses
- L27:1430 weakly suppresses
- L24:3003 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
- batim
- 的时代
- ibili
- 相亲
- bearer
- iddles
- 仆
- 非常高
- ating
- Bender
Suppresses
- area
- Area
- areas
- area
- Area
- _area
- -area
- areas
- (area
- 面积