the word 'age' and 'how many'
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. 5 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them 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 word "age" in phrases like "voting age", "age 13" or "age-related decline", and a little on "how many". The steering outputs don't show a clear link to it.
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.
-
graduating, she practiced in Germany where she also supported women doctors, helping to establish a women’s clinic and hospital.<|endoftext|>It's Hiking Season, But How Many Calories Are You Really Burning?⏎When the weather gets colder, one of the most
-
Dr. Nuha El Sayed: Thank you for having me.⏎So, today I would love to pick your brain about insulin resistance because it affects now how many tens of millions of people in our country, and then even more across the globe
-
aims to build the capacity and commitment of young Canadians to participate in their democracy. Student Vote's flagship program is a parallel election for students under the voting age, coinciding with official election periods. The program combines in-class learning, family
-
age 13 their minds and bodies are going through radical changes, which makes it difficult for them to focus on anything else. This might be too early an age to begin driving lessons, but
-
age, the authors report, while those who moved to the mainland — and consumed about 10 times less cocoa than their island-dwelling peers — showed age-related increases in blood
-
applied. A total of twenty nine depression patients in Sakura room were studied. Dependent t-test was applied to analize the data. The mayority of the age respondent between 16-24 years old. The male was fewer 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: Like a person: age in number of years: 2023-07-10, 20:00:00-07:00, 2023:04:0
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
regular picking: peas, beans, tomatoes, leafy crops, courgettes⏎3. WHEN? (see crop chart)⏎Right time: temperature, how long til harvest? Seed packets will tell you when to sow &
-
should be called getVolumeOfSphere(), take one parameter called radius and apply the relevant formula to calculate and return the volume of a sphere based on
-
efficiency is affected by angle of incline, characteristics of the specific bulk material, type of screw conveyor trough and screw pitch. Gkindus recommends designing screw conveyors using
-
tendency of different migrant groups to take out citizenship and to measure the size of groups eligible to vote. The data are useful cross-classified with birthplace, year of arrival in Australia and age data.⏎This page last updated 27 July
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
- L2:324 suppresses 'less' in 'colorless'
- L4:379 excites
- L4:340 suppresses
- L4:1604 excites
- L4:299 excites
- L4:2497 suppresses
- L4:1065 excites
- L3:165 suppresses
- L4:3007 excites
- L0:2162 weakly suppresses
- L4:464 weakly suppresses
- L3:728 weakly excites
- L4:660 weakly excites
- L1:1529 weakly suppresses
- L4:2251 weakly excites
- L4:1385 weakly suppresses
- L3:354 weakly suppresses
- L3:2129 weakly suppresses
- L4:913 weakly excites
- L2:852 weakly suppresses
- L4:980 weakly excites
- L4:1217 weakly suppresses
- L3:1996 weakly suppresses
- L4:1791 weakly suppresses
Feeds later layers
- L25:1055 excites
- L8:1672 weakly suppresses
- L20:1891 weakly excites
- L7:1296 weakly excites
- L10:2887 weakly excites
- L7:423 weakly suppresses
- L8:392 weakly suppresses
- L8:382 weakly excites
- L6:686 weakly suppresses
- L12:2347 weakly suppresses
- L10:1197 weakly suppresses
- L14:3009 weakly suppresses
- L20:819 weakly excites
- L15:2244 weakly excites
- L17:1834 weakly suppresses
- L6:205 weakly excites
- L16:1091 weakly suppresses
- L8:2046 weakly excites
- L14:2302 weakly suppresses
- L22:1881 weakly suppresses
- L25:1645 weakly excites
- L13:1407 weakly excites
- L20:1844 weakly excites
- L18:1688 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
- !"; ⏎
- adero
- ugi
- ... ⏎
- elic
- ... ⏎
- 痘
- Magick
- edList
- íveis
Suppresses
- vant
- 网讯
- ivant
- ソン
- (insert
- це
- inch
- OkHttpClient
- ˣ
- lop