'sin' and 'low' in technical text
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. 23 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 a mix of words in science text, most often 'sin' in trig formulas, 'low' before a level, and 'yst' in words like crystallized. The steering outputs do not show a clear theme, so its job is hard to pin down.
Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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.
-
ital hypothyroidism as part of standard screening tests. These tests detect almost all cases of congenital hypothyroidism. Doctors check blood levels of⏎Low levels of T4 and high levels of TSH suggest a diagnosis of congenital
-
power is then written as⏎Here the average of sin2 ωt over a cycle is 1/2 and that of sin ω t cos ωt is zero. Substituting these values,
-
of Amorphous panels is the low cost and flexibility of application. See also single-crystal silicon and polycrystalline silicon.⏎Angle of Incidence — The angle that a ray of sun
-
inhaled are generally used in the form of metered aerosols.⏎in the treatment of COPD apply expectorants and mucolytics ("Acetylcysteine", "Bromhexine",
-
AC sinusoidal signal which is characterized by its voltage V(t)=Vmax×sin(ωt+ΦV) and current I(t)=Imax×sin(ωt+ΦI) where Vmax, Imax are the peak
-
8, Martinus Beijerinck⏎showed that a filtered, bacteria⏎-free culture medium still contained the infectious agent. Wendell Meredith Stanley⏎crystallized the virus in 1935⏎and showed that it remains
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: how to include a single line in a line-include-er/er/er/er/er/er/er/er/er/er/er/er/er/er/er/er/er/er/
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
were tested on the basis of clinical suspicion or risky exposure. A human case of brucellosis was considered confirmed if results of serologic tests, such as ELISA or complement fixation test, were positive, in accordance with the World Health
-
ad (19th of August, in leap years 18th August) is the national holiday. Remembering that Afghanistan got its independence in 1919
-
determines your blood type. There are two main blood groups, the ABO group and the Rh group. The ABO group is made up of A, B, AB, or O blood types, and the Rh group is either positive or negative
-
British Straits Settlements in 1826. Occupied by the Japanese during World War II, Singapore became independent from the United Kingdom in 1963 and united with other former British territories to form
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
- L3:442 excites
- L21:2070 weakly suppresses
- L5:2398 weakly suppresses
- L17:578 weakly suppresses
- L7:2257 weakly excites
- L22:3029 weakly excites
- L5:692 weakly suppresses
- L14:1095 weakly suppresses
- L22:2181 weakly suppresses
- L14:1895 weakly suppresses
- L7:1261 weakly excites
- L10:492 weakly suppresses
- L8:1459 weakly suppresses
- L18:1778 weakly excites
- L22:3034 weakly suppresses
- L1:1558 weakly suppresses
- L7:1232 weakly suppresses
- L3:731 weakly excites
- L6:3024 weakly excites
- L16:2823 weakly suppresses
- L7:991 weakly excites
- L13:708 weakly excites
- L6:2083 weakly suppresses
- L22:1400 weakly suppresses
Feeds later layers
- L27:2281 weakly suppresses
- L27:1631 weakly excites
- L26:1137 weakly excites
- L27:1376 weakly excites
- L27:1951 weakly suppresses
- L27:1817 weakly suppresses
- L25:2042 weakly suppresses
- L25:1339 weakly excites
- L26:2688 weakly suppresses
- L25:2886 weakly suppresses
- L27:1825 weakly suppresses
- L26:1426 weakly suppresses
- L25:1068 weakly suppresses
- L24:2908 weakly excites
- L27:1721 weakly excites
- L24:1711 weakly excites
- L25:1180 weakly suppresses
- L24:2859 weakly suppresses
- L25:1778 weakly suppresses
- L24:181 weakly suppresses
- L24:1166 weakly suppresses
- L27:1276 weakly suppresses
- L24:609 weakly excites
- L26:49 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
- FileMode
- setType
- Fl
- CAS
- Lib
- lib
- vacc
- 永
- ┬
- (lib
Suppresses
- irit
- /Product
- leting
- eker
- oulder
- ество
- rude
- unched
- 鳌
- lexible