the digit 9 at a fragment start
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 reacts to the digit 9 when it is the first character of a piece of text, often the start of a year or number. Turning it up produced no obvious theme in the output.
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.
-
9 testing relying on micro/nano technology.⏎- Micro/nano technologies for drug analysis
-
932. The hotel was used as a seminary for the Society of Jesus
-
9 January 1875 to Lizzie Mark and her husband, Henry Benjamin
-
92 she passed the university junior scholarship examination.⏎In 1893 Eth
-
9-0.965). Children were more vulnerable to rotavirus (p
-
990 to 2002. It was the first effigy to
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, and I have been thinking about lately, and I have been thinking about lately. Maybe you have been thinking about lately, and I have been thinking about lately. So, the problem is that I have
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
most developed economies, is home to The Beatles, the birthplace of rock and pop music. England is also a site of numerous ancient Stonehenge, Baths of Stonehenge and thousands of years-old castles and churches. The Welsh valleys
-
with Washington Irving and Samuel Taylor Coleridge.⏎Returning to England, he was very influenced by romantic artists like
-
toplasm of these cells majorly is covered by the granules, excluding the bottom. The basal portion comprise of nucleus, mitochondria, endoplasmic reticulum and the other organelles. Microvilli are projected by
-
his personal diary. He was not the only US President to grow hemp however, Thomas Jefferson, James Madison, James Monroe, Andrew Jackson, Zachary Taylor, and Franklin Pierce were all reported to have sown the hemp seed!
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
- L1:1619 excites the very first token '0'
- L1:110 excites
- L1:1662 excites
- L1:81 suppresses
- L1:460 excites
- L21:1884 weakly suppresses
- L1:1337 weakly excites
- L2:60 weakly excites
- L1:218 weakly excites
- L1:437 weakly excites the digit 9 at a fragment start
- L4:278 weakly excites
- L26:648 weakly excites
- L1:1321 weakly excites
- L22:1654 weakly excites
- L1:2147 weakly excites
- L1:1262 weakly excites
- L17:2834 weakly excites
- L1:1431 weakly excites
- L26:21 weakly excites
- L17:2268 weakly excites
- L10:2459 weakly excites
- L2:128 weakly excites
- L17:1433 weakly excites
- L19:706 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
- variable
- moments
- outcomes
- solid
- long
- best
- moment
- )[-
- aff
- (L
Suppresses
- leases
- Submission
- äß
- _submission
- arkan
- utron
- mdb
- lder
- iedade
- 入口