"years" after a distance or time number
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. 13 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 word "years" when it follows a number in phrases like "12 million light years" or "Fifty years ago". Turned up, the output did not clearly show this theme, so the effect is hard to read.
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.
-
Earth. M81 is a spiral galaxy about 12 million light years away that is both relatively large in the sky and bright, making it a frequent
-
Only five years ago, it was uncovered that Patagonia was filling jackets using suppliers who pl
-
stars. Given how big it is in the image, and accounting for the magnifying effects of the gravitational lens, it’s less than 500 light years across, comparable to big, bright nebulae in nearby galaxies.⏎Here’s
-
The same techniques that predict planetary motions here at home in the solar system also describe the gravitational process that assembles thousands of galaxies into filaments a hundred million light-years long.⏎A major new series of cosmological simulations is now beginning to release its
-
such as SDSS J122952.66+112227.8. This star is 55 million light years away. The light we see today, comes from a time 10 million
-
, Ky. – July 16, 1969, is a historic day in American history, but also the history of civilization.⏎Fifty years ago, NASA launched the Apollo 11 mission, sending astronauts to the moon
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 make it simple and just plain. The focus is on the mind. The focus is just plain. Now let's analyze it. The focus is on the mind. The focus is just plain. Now let's's analyze it.
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
وان مقاله [English]⏎نویسنده [English]چکیده [English]⏎The author in this article has chosen all Imam Ali’s (A.S) speeches in Nahj-Albalagheh which are specifically about
-
:46 [Source: ICIS news]⏎By Nigel Davis⏎LONDON (ICIS news)--The carbon reduction goals likely to be set at the Copenhagen UN
-
as the ways enslaved peoples fought for liberation, workers’ rights movements, the intimate relationships of families, and the power struggles of monarchs and presidents, to name only a few.⏎History faculty are committed to fostering a classroom environment which is welcoming and
-
nothing until it is achieved, and with the initial focus often projected at how to pass the entry exams, the aftermath of admission has been a nightmare for thousands if not millions of students.⏎English is one the most widely spoken international languages and many would
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
- L12:459 excites
- L12:2272 excites
- L11:1384 suppresses
- L11:294 weakly excites
- L11:2093 weakly suppresses
- L9:910 weakly excites
- L11:1723 weakly suppresses
- L9:401 weakly excites
- L11:1210 weakly suppresses
- L8:2506 weakly suppresses
- L12:267 weakly excites
- L8:127 weakly excites
- L8:119 weakly suppresses
- L12:2963 weakly excites
- L9:288 weakly excites
- L6:85 weakly suppresses
- L8:280 weakly suppresses
- L9:744 weakly suppresses 'a' after a preposition
- L9:1027 weakly suppresses 'to' or 'on' after 'was' or 'were' (planned future action)
- L9:970 weakly suppresses
- L11:119 weakly excites
- L7:14 weakly excites
- L12:1303 weakly excites
- L10:1168 weakly excites
Feeds later layers
- L14:1430 suppresses
- L15:481 weakly suppresses
- L16:2366 weakly excites
- L14:686 weakly suppresses
- L18:70 weakly suppresses
- L14:2344 weakly suppresses
- L25:376 weakly excites
- L14:542 weakly suppresses
- L21:908 weakly excites
- L16:395 weakly suppresses
- L18:941 weakly excites
- L14:2636 weakly suppresses
- L16:222 weakly suppresses
- L16:2625 weakly suppresses
- L23:2449 weakly suppresses
- L15:167 weakly suppresses
- L19:423 weakly suppresses words that set up a following clause
- L25:591 weakly excites
- L16:672 weakly suppresses
- L16:2307 weakly suppresses
- L19:958 weakly suppresses
- L16:1872 weakly excites
- L17:1343 weakly excites
- L17:384 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
- Corinthians
- 真实的
- (be
- begging
- ldb
- ships
- 孛
- 兹
- 的真实性
- Coupons
Suppresses
- iş
- riteln
- íst
- ní
- plash
- trance
- üş
- üss
- angkan
- sizei