"x" or "by" between two 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. 8 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 strongly on the "x" or "by" in measurements like "20 x 17 cm", "30 x 30" or "2 miles wide by 6 miles long". It is one of the clearest neurons in this set: it reacts to the middle of a size or dimension description. Steering at higher strength produced number-heavy text in some cases, which fits loosely.
Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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.
-
how it deals with death. Published by Edition Patrick Frey (Zurich).⏎250 pages, 20 x 17 cm, softcover, Edition Patrick Frey (Zurich
-
For example, you may find a threaded rod measuring M8-1.25 x 1 Meter. The M8 (8mm) is the rod's diameter
-
future town. It will be massive: 2 miles wide by 6 miles long and up to 10,000 homes.
-
("Nanoholes As Nanochannels: Flow-through Plasmonic Sensing"), researchers at the University of British Columbia created a grid of 30 x 30 flow-through nanoscale holes to create a highly responsive sensor system
-
Northumberland).⏎The fort itself covers an area of about 4½ acres (c.1.8 ha), measuring 515 feet north-south by 385 feet east-west (157 x 117
-
, 1649.⏎Etching and engraving on laid paper, with margins; letterpress text on verso; platemark: 422 x 531 mm; total: 518 x 63
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: is the problem, or? I am, or? I, or? I, or? I, or? I, or? I, or? I, or? I, or? I, or? I, or? I
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
not died? (9) They had returned the home? (10)You had finished this book?⏎Past Perfect Tense Negative⏎Subject + had not + verb 3rd from + object.⏎Past Perfect Interrogative Negative (Examples
-
is expected to increase, which is a growing public health concern.⏎|Status||Published - 2014 jan. 1|⏎- Folkhälsovet
-
door size⏎Place the fully charged battery in an environment of 55°C±5°C. After the surface temperature of the battery reaches 55°C±5°C, leave it for 30 minutes, then connect the positive and
-
However, since it is a function instead of syntax, it can be partially applied and passed around to higher order functions, like so:⏎ghci> :m + Data.Bool.Extras⏎ghci> let yesOrNo = bool
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
- L4:799 suppresses
- L4:760 excites
- L5:2079 excites
- L7:42 excites
- L7:1000 weakly excites
- L5:640 weakly suppresses
- L6:968 weakly suppresses
- L7:1664 weakly excites
- L1:360 weakly suppresses
- L4:717 weakly excites
- L3:726 weakly suppresses
- L6:2682 weakly excites
- L6:1084 weakly suppresses
- L2:112 weakly excites
- L1:2120 weakly excites
- L2:2009 weakly suppresses
- L5:117 weakly excites
- L1:660 weakly excites 'with' in 'along with'
- L3:1668 weakly suppresses
- L7:2962 weakly suppresses
- L6:604 weakly suppresses
- L3:2031 weakly excites
- L7:2320 weakly suppresses
- L7:23 weakly suppresses
Feeds later layers
- L12:51 suppresses
- L9:878 excites
- L14:706 excites
- L16:1074 excites
- L18:1857 excites
- L9:651 suppresses
- L16:314 excites
- L9:326 excites
- L12:336 suppresses
- L12:2548 excites
- L18:1270 excites
- L18:870 excites
- L9:212 suppresses
- L16:1999 excites
- L19:1823 excites 'and' inside organisation or paired names
- L9:1047 excites
- L12:1095 excites
- L10:1898 excites
- L14:917 excites
- L13:27 suppresses
- L11:682 excites gap before a second number
- L15:1162 excites
- L14:1054 excites
- L10:272 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
- eten
- >'+⏎
- ">'+
- Department
- ">'+⏎
- <?⏎
- >"+⏎
- "").
- ultureInfo
- "\⏎
Suppresses
- umbai
- 滂
- amble
- like
- ович
- -know
- gressive
- wrapper
- /auto
- egade