"of" after a time or place noun
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. 6 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 weakly on the word "of" after a time phrase like "in the year of", "the time of" or "the month of". It is a weak signal, and turning it up gave unrelated text, so the steering doesn't show the same idea.
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.
-
the result of an intense source of energy as the body passed through the cloth, perhaps at the moment of resurrection,” reported The Guardian.⏎Believed to be shroud of Jesus⏎The
-
. The title of the story is kept by the name of the village only. So the title truly signifies the story.⏎Born on November 8 in the year of 1908 Raja Rao was a famous writer. The place was
-
-old, who is an Electrical Engineer from Usha Rama College of Engineering and Technology, Vijayawada.⏎The product was to be launched in the month of May<|endoftext|>Scientists saw the Earth from a new perspective on 24th October
-
environment.⏎Despite that good news, the study further concluded that deaths in the year of the study publication would be in excess of 300,000.
-
hearts or arteries preserved enough to study. Of those, nine had evidence of blockage from atherosclerosis. “This disease has been around since before the time of Moses,” said Randall Thompson
-
received his Undergraduate and Master 's degrees. After receiving his degrees he became a faithful lawyer and was chosen to help draft the Declaration of Independence in the year of 1774 while also serving on the first Continental Congress. In the
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: I want to have a system where the user can use a system where the following equation is given: (x + y + z) = (x + 1) * y * z. What is the name of the system that I
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
gdalin. Amygdalin is a plant compound that contains sugar and produces hydrogen cyanide. Hydrogen cyanide is thought to be the main anticancer compound formed from laetrile when broken
-
alin. Amygdalin is a chemical ingredient found in the pits of many fruits, raw nuts, and plants (see Question 1).⏎- Hydrogen cyanide is thought to be the main anticancer compound formed from laetrile when broken
-
of concern have been rising in atmospheric concentrations over the last two hundred or so years, since the beginning of the industrial age.⏎Those involved are methane, nitrous oxide, fluorocarbons, sulphur hexafluor
-
simply our way of life today: carbon dioxide, methane, nitrous oxide, hydrofluorocarbons, perfluoroc<|endoftext|>Nowadays,
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
- L5:96 weakly suppresses
- L4:330 weakly suppresses
- L4:2049 weakly suppresses
- L1:369 weakly excites
- L1:2648 weakly excites
- L2:2965 weakly suppresses
- L3:2787 weakly excites
- L1:1055 weakly excites
- L0:2893 weakly suppresses
- L3:987 weakly suppresses
- L4:2632 weakly excites
- L0:2928 weakly suppresses
- L2:388 weakly suppresses
- L2:2498 weakly excites
- L2:1103 weakly excites
- L5:1075 weakly excites
- L3:348 weakly excites
- L4:1801 weakly excites
- L2:1899 weakly excites
- L1:1229 weakly excites
- L3:1877 weakly excites
- L4:52 weakly excites
- L4:2798 weakly suppresses
- L2:1469 weakly excites
Feeds later layers
- L7:1882 excites
- L8:333 excites
- L7:2687 weakly excites
- L18:217 weakly suppresses
- L25:669 weakly excites
- L11:157 weakly excites
- L21:2251 weakly excites
- L19:776 weakly excites
- L10:2838 weakly excites
- L16:1422 weakly excites
- L25:1695 weakly excites
- L20:585 weakly excites
- L8:2051 weakly excites
- L25:756 weakly excites
- L27:610 weakly suppresses
- L24:1526 weakly excites
- L27:1448 weakly suppresses
- L10:1148 weakly excites
- L24:1716 weakly excites
- L24:713 weakly excites
- L19:1691 weakly suppresses 'the' inside a phrase, mid-sentence
- L24:1693 weakly excites
- L22:43 weakly suppresses
- L9:758 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
- Wien
- Natal
- conver
- .AddDays
- -Pacific
- noon
- 收敛
- 氚
- Advent
- ather
Suppresses
- elling
- этому
- icers
- 依法追究
- 一堆
- legate
- гляд
- 四大
- 一枚
- 两张