document boundary 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. 12 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 most on the invisible 'end of text' marker that separates one web page or document from the next. It also fires weakly on a few words that start a new idea. Turned up, the model's output became repetitive and unfocused.
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.
-
are checked to verify they meet the required specifications.⏎<|endoftext|>RASPBERRY Pi - LEARNING HOW TO MAKE YOUR ROBOT WORK⏎
-
firstname.lastname@example.org.<|endoftext|>The Musical Representation⏎How human musical experience emerges from the audition of organized tones is
-
Hospitals use it to access patient records and share test results. Governments use it for elections, services and disaster alerts. The global financial system—bank transfers,
-
ycetica<|endoftext|>Adding Fractions Level 3⏎The 5-step planIn Adding Fractions
-
to give updated information once medical findings become available.<|endoftext|>History of Women Cheesemakers⏎Throughout human history, women were the primary
-
on the book "Dots and Dashes: Adventures at the Dawn of the Wireless Age" is at http://vubooks.com<|endoftext|>Child health nursing is designed to help the students to develop and understand the concept 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: the more I have been thinking about, the more I have been thinking about it. So, I have to try to find something in it, and so on. In other words, the more I have been thinking about, the more I
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
rih B. Approach to family assessment and intervention. Natl Med J India 2017;30:27
-
the Aurora Borealis. It contained standard scientific instruments.⏎But things - almost - went horribly wrong.⏎Routine
-
the discs that cushion the bones of the spine (the vertebrae). To treat degenerative disc disease, doctors usually begin with conservative (nonsurgical
-
Canada, by the British North America Act (July 1, 1867). The province of Canada is divided into Ontario and Quebec. The United States of America
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
- L10:369 strongly suppresses
- L9:1033 suppresses unclear: end-of-text markers and word fragments
- L7:1053 suppresses
- L11:784 suppresses
- L11:607 excites
- L7:1130 excites
- L10:762 suppresses
- L4:1411 excites
- L8:1515 suppresses
- L6:927 suppresses
- L11:1143 suppresses
- L9:1131 suppresses
- L11:203 suppresses
- L4:89 suppresses
- L10:258 weakly excites
- L0:1128 weakly suppresses
- L7:911 weakly suppresses
- L10:1846 weakly excites
- L8:489 weakly excites
- L0:999 weakly suppresses
- L5:442 weakly suppresses
- L1:981 weakly excites
- L0:1867 weakly suppresses
- L7:746 weakly excites
Feeds later layers
- L13:1481 suppresses
- L15:606 suppresses
- L13:201 suppresses
- L17:320 suppresses
- L16:539 suppresses
- L18:53 suppresses
- L15:1814 weakly excites
- L14:969 weakly suppresses
- L15:794 weakly suppresses
- L15:2218 weakly excites
- L13:1681 weakly excites
- L13:1036 weakly excites
- L15:1034 weakly excites
- L16:2398 weakly excites
- L13:3032 weakly excites
- L14:1710 weakly suppresses
- L14:2982 weakly excites
- L13:1316 weakly suppresses
- L19:159 weakly suppresses
- L17:2031 weakly excites
- L19:393 weakly suppresses
- L19:55 weakly suppresses
- L22:21 weakly suppresses
- L18:44 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
- genie
- risen
- resses
- bas
- odes
- encing
- 觐
- ulling
- Genesis
- ouver
Suppresses
- 脂
- -badge
- herself
- otive
- alary
- enda
- Câmara
- ạt
- .HTML
- otos