unclear: end of document token
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. 19 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
The neuron's strongest reactions are around the marker that separates one web page from the next in the training text, which the model sees as a hard stop. Its large reactions are negative, and it does nothing clear when turned up.
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.
-
. It will contain an appropriate error⏎message for the user.⏎Upon success, this is the file handle returned by⏎``urllib2.urlopen``. Per urllib2, it has an ``info()`` method⏎containing things like headers
-
Increases Cardiovascular Risks⏎Led by James L. Vacek, M.D., a team of researchers at the University of Kansas Medical Center undertook a<|endoftext|>The Tenderlett lancet was made for single use in a lancing device.
-
the China site (in Chinese or English) for best site performance. Other MathWorks country sites are not optimized for visits from your location.⏎Contact your local office<|endoftext|>It’s PrAACtically Martin Luther King Jr Day⏎Looking for activities and
-
, or whether it’s heavy metals, or whether it’s nitrogen and phosphorus – will make it to the sea. And that means that probably water quality will<|endoftext|>Harlem riot of 1935⏎The Harlem riot of 1
-
it encountered the overwhelming gravitational force of the much larger mass. This force wrenched luminous stars from the small galaxy. Over billions of years of evolution, the<|endoftext|>Duplicating Data Elements⏎Using keywords and synonyms is one way to ensure that
-
the China site (in Chinese or English) for best site performance. Other MathWorks country sites are not optimized for visits from your location.⏎Contact your local office<|endoftext|>The point in the Moon's orbit where it lies diametrically⏎opposite
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: the first time I met with the first #1 #2 #3 #4 #1 #4 #1 #1 #1 #1 #1 #1 #1
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
'S', then click at the end of the word and type<|endoftext|>Mallorca’s popular tourist railway, the ‘Orange Express,’ has entered its
-
and Ragusa.<|endoftext|>Yellow fever is a disease carried by female mosquitoes. The species of mosquito that carry
-
running. Slow to<|endoftext|>South-eastern Australia is one of the most bushfire-prone areas in the
-
comment.<|endoftext|>100 Chemistry Questions and Answers 8211; Basic Chemistrynbsp
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
- L9:18 suppresses text boundary after a document break
- L9:114 suppresses
- L10:32 excites
- L17:644 suppresses
- L8:341 suppresses
- L18:109 suppresses
- L3:163 excites
- L3:5 excites
- L4:382 suppresses
- L17:1334 excites
- L18:555 excites end-of-text marker between documents
- L9:948 excites
- L9:49 suppresses
- L7:467 excites
- L1:1035 weakly excites
- L2:1092 weakly excites
- L5:2 weakly suppresses
- L7:453 weakly excites
- L14:67 weakly suppresses
- L15:570 weakly suppresses
- L10:1772 weakly excites
- L2:322 weakly excites
- L9:1130 weakly suppresses
- L5:1007 weakly excites
Feeds later layers
- L21:282 suppresses
- L20:96 excites
- L25:436 suppresses
- L22:221 suppresses period ending a contact or help sentence
- L26:1283 suppresses
- L21:955 suppresses
- L22:191 suppresses
- L27:1300 suppresses
- L27:256 suppresses
- L26:120 suppresses
- L26:601 suppresses
- L26:228 suppresses
- L22:230 weakly suppresses
- L21:652 weakly suppresses
- L22:2447 weakly excites
- L23:813 weakly excites
- L20:255 weakly suppresses
- L20:116 weakly suppresses
- L27:118 weakly suppresses
- L22:640 weakly suppresses
- L24:161 weakly suppresses
- L22:2374 weakly suppresses
- L21:633 weakly suppresses
- L26:59 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
- anced
- …”
- …)
- +]
- arks
- ographers
- \"]
- =").
- ”).
- :\"
Suppresses
- import
- ⏎ ⏎ ⏎
- ⏎⏎⏎
- https
- ⏎
- ⏎ ⏎
- ⏎ ⏎ ⏎ ⏎
- ⏎⏎⏎
- ⏎⏎⏎⏎