the digit 1 in age ranges
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. 25 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 responds to the number 1 when it starts an age or a range of ages, such as "ages 18-24" or "aged 18 and older". Its opposite side fires on the 1 in powers of ten like "10^8", and turning it down made the model write endless zeros.
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 two most dangerous activities your teen will ever participate in. Vehicle crashes are the number one cause of death for young adults ages 14-24!⏎Sign up for a free on-line parent program provided to
-
high enough to mean that 20% of all deaths in this age group in 2016 involved opioids.⏎Among those between the ages of 15 and 24, the report adds, the nearly 3,0
-
current study, healthy adults (age 18-35) read short discourse scenarios for comprehension while monitoring for the presence
-
in the late fall through winter. Some may experience SAD in the summer, but not as frequently. SAD is more common in young people (ages 18-30), particularly women. According to the Cleveland Clinic, higher risk
-
Cancer Society researchers led by Vilma Cokkinides, Ph.D., analyzed survey results from 4756 smokers (aged 18 and older) who visited a healthcare provider within the past year. All were
-
, year and sponsor across US cable networks and programs. Researchers calculated ad exposure for youth ages 12 to 17 years of age and young adults 18 to 24 years old.⏎The research showed that more than
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:000000000000000000000000000000000000000000000000
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
= 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
-
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.
-
-Boltzmann formula:⏎E = εσT4⏎where ε=emissivity, a material property; σ = 5.67×10-8 ; T = temperature in Kelvin (absolute
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
- L17:512 strongly suppresses
- L18:499 strongly excites
- L20:416 strongly excites the "1" in years 191x
- L3:244 suppresses
- L0:871 excites
- L4:247 suppresses
- L10:848 suppresses
- L19:720 excites
- L4:1401 excites unclear: letter fragments inside names
- L22:332 suppresses
- L21:518 excites
- L2:2029 suppresses
- L22:2733 suppresses
- L2:1866 excites
- L13:620 weakly excites
- L16:113 weakly suppresses
- L21:1031 weakly suppresses
- L5:559 weakly suppresses
- L20:287 weakly suppresses
- L10:722 weakly suppresses
- L21:271 weakly suppresses
- L2:1724 weakly excites
- L18:81 weakly suppresses
- L2:1699 weakly excites
Feeds later layers
- L26:413 suppresses
- L27:1153 suppresses
- L27:1196 suppresses
- L26:848 suppresses
- L27:153 excites the spot just before a historical number
- L27:205 suppresses
- L26:737 suppresses
- L27:862 suppresses
- L26:121 suppresses
- L27:396 suppresses
- L26:233 excites
- L27:238 suppresses
- L26:1461 suppresses
- L27:530 weakly excites
- L26:216 weakly suppresses blank slot before a year or number
- L27:1329 weakly excites
- L27:241 weakly excites
- L26:1065 weakly excites
- L27:1087 weakly excites
- L27:132 weakly suppresses
- L27:2659 weakly excites
- L26:598 weakly suppresses spot just before a year or century
- L26:905 weakly excites
- L27:902 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
- HIP
- ��
- ourney
- logg
- developers
- 力
- เต
- tees
- Dere
- rack
Suppresses
- 0
- 0
- ₀
- ǒ
- 零
- OUSE
- од
- ouse
- če
- ор