digit '9' inside years and numbers
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. 0 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 a digit inside a year or number, especially the 9 in years like 1993 or 1999. It is mostly about digits in dates and codes. Turning it up gave unrelated text, so the steering does not show a clear theme.
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.
-
via Getty Images<|endoftext|>Because RNA is such an important part of how the cell builds proteins, some scientists speculated that this was its only job. In 1993, however, Victor Ambros, Rosalind Lee, and Rhonda
-
cutting condition can be maintained on the entire path.<|endoftext|>OHIO DEPARTMENT OF AGRICULTURE⏎Division of Plant Industry – Apiculture⏎8995 East Main Street⏎Reynoldsburg, Ohio 4306
-
and teachers?⏎According to the Montana Common Core Standards<|endoftext|>Copyright © 1995 Brother Eric Vogel, St. Mary's College⏎Balsamor
-
a floral shop.⏎Even more influential of Liberia<|endoftext|>Oxford University Press⏎2017 | 200pp | £7.99⏎Organic chemistry is one of the most exciting areas of modern science, with
-
the past 20 years, Florida has been one of the leaders among the 50 states in the standards and accountability movement. In 1999, continuing in this tradition, the Florida Legislature adopted the A+ Plan for Education
-
, your<|endoftext|>Longwall dust control-where we are and where we need to go in the 21st century.⏎Longwall USA, 1998 Jun; :123-138⏎Although the number of
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 have been thinking about lately. I have been thinking about lately. I have been thinking about lately. I have been thinking about lately. I have been thinking about lately. I have been thinking about lately. I have been thinking about lately
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
CT requires strong<|endoftext|>90 LettersLanguage brain teasers are those that involve the English language. You need to
-
School of Mathematics Doctoral Training Grant funding - please contact us for more information.<|endoftext|>História (São Paulo)⏎On-line version ISSN 1980-4369⏎DOMINGUES, Beatriz Helena. The guar
-
scholar Jean O’Brien: Drop Dead<|endoftext|>Note: Supplemental materials are not guaranteed with Rental or Used book purchases.⏎- ISBN: 9781908049490 | 1908049
-
$800 million dollars a day-$13,000 a second-to spend
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
None: this is the first layer.
Feeds later layers
- L4:2938 excites
- L6:30 excites
- L4:308 weakly excites
- L1:1148 weakly suppresses
- L16:353 weakly excites
- L1:18 weakly suppresses
- L2:1594 weakly excites the number '100' in numerals
- L7:1793 weakly suppresses
- L18:196 weakly excites
- L2:400 weakly excites
- L26:887 weakly excites
- L6:2631 weakly excites
- L3:2880 weakly excites
- L2:2449 weakly excites
- L13:2655 weakly excites
- L22:2675 weakly excites
- L23:2465 weakly excites
- L16:917 weakly suppresses 'of' inside organization names
- L10:2038 weakly excites
- L10:757 weakly suppresses
- L26:2647 weakly suppresses
- L1:5 weakly suppresses
- L12:1747 weakly suppresses
- L5:332 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
- WithValue
- ECH
- atural
- aments
- ETA
- .toHexString
- ),
- criptors
- ograf
- 氾
Suppresses
- allis
- ้ง
- yne
- link
- null
- uro
- core
- lip
- Stations
- rophic