the digit 1 in powers of ten
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 digit 1 when it starts a power of ten, as in 3 x 10^8 or 10^-11, and the positive side is about big or tiny numbers written in scientific notation. Turning it up made the model repeat number words like 'ten' and 'nine', which matches what it was tracking.
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.
-
= 0, and mℓ = 0 is where σ = r/a0 and a0 is the Bohr radius (5.29 × 10-11 m). Calculate the position of the nodes for this wave function
-
substance of one density into a substance of different density, the speed of light changes and bending of light occurs.⏎In air, light travels at 3 x 108 m/sec but in water, it travels only about 2.2
-
-Boltzmann formula:⏎E = εσT4⏎where ε=emissivity, a material property; σ = 5.67×10-8 ; T = temperature in Kelvin (absolute
-
water than in air, water is said to have a greater optical density than air.⏎Substance Speed of light (m/sec)⏎Vacuum 3 x 108⏎Air 3 x 108⏎Water 2.
-
small when written as 52!, but it can also be written in scientific notation as 8.065817517 × 10^67. If I attempt to write it using normal numerals,
-
2.4 \times 100 \div 3.2$.⏎You could, of course,
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: setting up a set of settings setportportportportportportportportportportportportportportportportportportportportportportportportportportportportportportportportportportportportportportportportport
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
strains⏎Klebsiella resistant strains have been recorded in USA with a roughly threefold increase in Chicago cases, Brazil with 15 deaths per 163 cases, quarantined individuals in Israel, United Kingdom and parts
-
a substance of one density into a substance of different density, the speed of light changes and bending of light occurs.⏎In air, light travels at 3 x 108 m/sec but in water, it travels only about 2.
-
iate ligament (ACL) is the most commonly injured ligament of the knee. Every year, there are around 30 cases of ACL injuries for every 100,000 people. If the ACL is torn, surgery
-
Re became Tefnut's father. According to one myth, Tefnut became angry with her father Re and ran off to Nubia. The god Thoth went to find her and conjoled her into returning to Egypt.⏎
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
- L24:1135 suppresses
- L24:2812 suppresses
- L24:482 suppresses inside Italian names and places
- L21:2868 excites
- L22:588 suppresses
- L23:352 suppresses
- L20:2963 suppresses
- L21:711 excites
- L25:744 excites
- L21:2552 excites
- L22:179 excites
- L16:91 suppresses
- L24:2346 excites
- L25:1895 suppresses first digit of a year like 2020
- L22:2871 weakly suppresses
- L24:831 weakly suppresses
- L22:1527 weakly excites
- L22:2744 weakly suppresses
- L20:221 weakly suppresses
- L22:1882 weakly excites
- L23:245 weakly suppresses
- L22:1492 weakly suppresses
- L24:106 weakly excites
- L6:449 weakly excites
Feeds later layers
- L27:328 suppresses
- L27:902 suppresses
- L27:180 excites
- L27:1522 suppresses
- L27:893 weakly excites
- L27:624 weakly suppresses
- L27:132 weakly suppresses
- L27:696 weakly excites
- L27:1741 weakly suppresses
- L27:2354 weakly excites
- L27:294 weakly excites
- L27:1944 weakly excites
- L27:849 weakly excites
- L27:2277 weakly excites
- L27:678 weakly excites
- L27:762 weakly excites
- L27:1316 weakly suppresses
- L27:1873 weakly excites
- L27:435 weakly excites
- L27:3049 weakly excites
- L27:847 weakly suppresses
- L27:439 weakly excites
- L27:2789 weakly excites
- L27:1202 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
- tenth
- Ten
- ten
- ninety
- ninth
- -nine
- nine
- _ten
- Ten
- ten
Suppresses
- orget
- setName
- сет
- 设
- 嫱
- ounter
- .setImage
- ulia
- setId
- setUp