'the' in 'hit in the body part' idioms
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 responds to the word 'the' in phrases like 'shot in the head', 'hit him in the back' or 'pat on the back', where a body part is about to be named. Turning it up did not produce any obvious theme in the generated text.
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.
-
Thomas, by Caravaggio (Michelangelo Merisi da, 1571–1610) which shows Thomas literally poking Jesus in the side to feel his wounds.⏎Pretty gross. It also depicts two other disciples looking
-
and Eighth circuits have all said, in so many words, no right exists to have a machine gun. You can add the U.S. Court of Appeals for the Ninth Circuit to the list.⏎In affirming the conviction of Alaskan Matthew Wayne
-
have the option of joining the expedition for two weeks.⏎HOW WILL YOU HELP⏎You<|endoftext|>Adding and formatting mathematical equations in your papers can be a real pain in the neck. This is especially the case with complicated expressions involving matrices, calculus, bin
-
in a railroad station, before shooting a second time.⏎One bullet only grazed Garfield’s shoulder but the other one, the fatal one, hit him in the back missing his spinal cord although it passed the first lumbar vertebra and lodged
-
his high school bathroom and shot himself in the head.⏎Megan Meier, 13, committed suicide after being harassed
-
Wentworth to Lyme.⏎- Louisa jumps off the Cobb steps and knocks herself unconscious on the ground.⏎- Captain Wentworth carries Louisa to the Harvilles’ home.⏎- The doctor says that Louisa will recover, but cannot be
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 want to be in the job, and I want to be in the role that I have been in is something that I want to be in is something that I have been in is something that I want to be in is something that I
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
into a robot.⏎“They didn’t actually know they could build it themselves,” said Santanna Lea, program director at the Ed Endres Boys & Girls Club.⏎She conceded she didn’t know much about robotics at the
-
engineering of plant clock systems, with powerful applications to agriculture.⏎(PRWeb UK) March 17, 2010 -- Researchers at the RIKEN Plant Science Center have clarified the function of three proteins that play a central role in
-
a teacher helped her students learn over the course of the year. Here are the facts that correct common misconceptions about the tool.<|endoftext|>Researchers at the RIKEN Plant Science Center have clarified the function of three proteins that play a central role in
-
for their male love interests.Continue Reading⏎The Sadie Hawkins tradition comes from the comic strip Li'l Abner in which women chase men around
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
- L10:120 strongly suppresses
- L10:1377 strongly suppresses
- L11:1168 strongly excites
- L9:662 strongly suppresses
- L10:321 strongly suppresses
- L10:1644 strongly excites
- L4:73 strongly suppresses
- L9:142 strongly excites
- L9:227 excites
- L9:2661 excites
- L3:344 excites
- L2:205 excites
- L1:740 suppresses
- L9:2198 excites
- L2:1439 excites
- L5:705 excites
- L3:299 excites
- L11:875 suppresses
- L10:365 suppresses
- L5:493 excites
- L5:153 suppresses
- L10:1162 excites
- L3:2559 suppresses
- L9:24 excites "the" before a named organisation or title
Feeds later layers
- L13:1698 strongly suppresses
- L14:826 suppresses
- L15:702 excites
- L14:343 excites
- L16:391 excites
- L15:309 excites
- L14:547 suppresses
- L19:1579 suppresses
- L14:2965 suppresses
- L19:2198 suppresses
- L16:1497 suppresses
- L19:2555 suppresses
- L17:1238 suppresses
- L15:668 suppresses
- L14:298 excites
- L13:906 suppresses
- L13:2727 suppresses
- L16:1715 suppresses
- L13:2634 suppresses
- L18:1770 suppresses
- L13:1365 suppresses
- L18:1326 suppresses
- L15:1645 suppresses
- L16:1643 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
- ках
- ,
- otland
- 新娘
- ,在
- ,默认
- thern
- аз
- 。
- oints
Suppresses
- -www
- aspect
- viso
- elevator
- 的身影
- utivo
- 身影
- 栏
- ería
- ibase