unclear: end-of-document markers
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. 25 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 is hard to pin down. Its strongest moments are the hidden 'end of text' markers that separate web pages, plus a few odd spots in code and lists. Steering did not reveal a clear theme.
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.
-
technically clean, and is not actually any cleaner than a human mouth.⏎Is saliva of dog harmful?⏎Recent case reports demonstrate the harms of Capnocytophaga canimorsus bacteria in dog saliva, which can cause
-
is used to count the numbers of characters in⏎a string, as well as the number of characters in the string variable⏎>>> len("a string") 8 >>> >>> s = "a long string" >>> >>> len(s)
-
Sep 2012 • 1 answer⏎Posted 3 Oct 2014 • 1 answer<|endoftext|>Choosing Whole Foods for a Healthier You⏎Written by Eric DeAngelis,
-
considered a simple problem, one that a fifth grader may find in a math book. What about the problems we face in adulthood? (read more here)<|endoftext|>Задание от гостя:Размещено 13.
-
can safely remove.⏎But remember that you should never cut off more than two-thirds of a root at any time.⏎Also,<|endoftext|>Title VII makes sexual discrimination, i.e., using gender as a reason for making
-
, u, t)$.⏎<|endoftext|>While many students view summer as a time to forget about school, it remains an
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: what is the more general term for a group of people who are themselves in their own interests and so are likely to favor their interests over other interests. What is the more general term for a group of people who are themselves in their own interests
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
Show them your budget and let them be a part of some of the money decisions in your household. Seeing how money works can be a big tool for a tween.⏎#8. Know That You Are Their
-
in the field from August 2 – August 20.<|endoftext|>Genetics as it applies to evolution, molecular biology, and medical aspects.⏎10 posts • Page 1 of 1⏎Hope somebody can help, I've got an
-
JBS10AMA. To make a password even more secure, or applicable for services that require other characters, add special characters around it, like # JBS10AMA#.⏎Here are some
-
means that the files are stored in a compressed format.⏎For Python Implementation⏎You can use OpenCV to manipulate the images.⏎pip3 install opencv-python numpy⏎Open
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
- L21:328 suppresses
- L22:259 excites
- L20:508 suppresses
- L8:51 suppresses
- L20:85 excites
- L21:341 excites
- L15:574 excites
- L22:2243 suppresses
- L20:843 suppresses
- L13:643 excites
- L3:38 suppresses
- L2:1499 suppresses
- L7:221 excites
- L16:496 weakly suppresses
- L19:389 weakly suppresses
- L10:754 weakly suppresses
- L0:2110 weakly excites
- L17:1837 weakly suppresses
- L20:213 weakly excites
- L19:439 weakly suppresses
- L4:243 weakly excites
- L5:2867 weakly suppresses
- L17:737 weakly suppresses
- L0:2476 weakly excites
Feeds later layers
- L26:855 strongly excites
- L27:952 strongly excites
- L27:1447 excites
- L26:612 suppresses end-of-text boundary tokens
- L27:619 suppresses
- L27:2159 excites
- L26:720 weakly suppresses 'New' in place names
- L27:865 weakly excites
- L27:614 weakly excites
- L27:1339 weakly excites
- L26:235 weakly suppresses
- L27:2277 weakly excites
- L26:611 weakly excites
- L27:2191 weakly suppresses
- L26:1072 weakly suppresses
- L27:2538 weakly suppresses
- L26:2218 weakly suppresses
- L27:1005 weakly excites
- L27:1848 weakly suppresses
- L26:2078 weakly suppresses
- L27:1456 weakly excites
- L27:2718 weakly suppresses
- L26:1391 weakly excites
- L26:609 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
- #
- #
- #a
- #"
- ,#
- #\
- #,
- #=
- #",
- #g
Suppresses
- //⏎
- //
- //
- //⏎
- //(
- //"
- ////
- )//
- //'
- //"