'Me' starting a surname
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 fires on the syllable 'Me' when it starts a name, such as Meier, Meade or Meinkoth, mostly surnames inside news and citation text. Steering it changed the topic a lot without a clear link to names, so the link to this trigger is only a guess.
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.
-
Want to Be President? is chock-full of interesting tidbits about our Presidents;...⏎Activities and Games, Extension Activities, Book Resources⏎Extension activity for Arthur Meets the President by Marc Brown.⏎Biography of former United States President George Herbert
-
described are explored⏎Footnote 1: adapted from Woodward-Kron, R., Thomson, E. & Meek, J. (2000) A text based guide to academic
-
called poults. A group of turkeys is called a flock.⏎The benefits of raising white turkeys are<|endoftext|>In this thoroughly researched history, Graydon Meints tells the fascinating story of the railroad’s arrival and development in Michigan. An
-
of his campaign, he had to protect 150 miles of<|endoftext|>Rethinking Teen Sex⏎By Tim Brady⏎A new study by assistant professor Ann Meier has made waves with its fresh look at adolescent sexuality⏎Almost half of American
-
(Charonia Research, 1996b; Charonia Research, 1996c; Hendler, 1995; Meinkoth, 1981)⏎This is a bottom dwelling species that
-
, it burned only the energy that was available from the environment, slowing down the ‘fire of life’ in times when resources became scarce,” said study co-author Meike Kohler, a paleobiologist at the Autonomous University of Barcelona, Spain
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:������������������������������������������������
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
timing circuit that gives precise and highly stable delays of time or oscillation. These types of ICs are very cheap and reliable in cost when we compared with the OP-Amp applications in the same areas. These ICs are used as an ast
-
next step is to determine the identifier for the drive you want to make an image of. In the terminal window, type in the command⏎sudo fdisk -l⏎Our 1GB drive is /<|endoftext|>During the period from 19
-
use any editor such as edit to write a C program.⏎C:> edit myprogram.c⏎- After written the program, to compile it, issue the command cl⏎C:> cl myprogram.c⏎- During the compilation and linking, my
-
agonists (Allbuterol/Salbutamol, Proventil, Ventolin) and theophylline
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
- L20:1257 excites
- L20:831 excites
- L13:44 suppresses inside web addresses and citations
- L10:155 excites
- L24:214 excites
- L15:184 excites
- L20:2046 excites
- L21:720 excites
- L12:158 suppresses
- L20:54 excites
- L11:223 excites
- L21:2711 excites
- L19:1685 suppresses
- L20:162 excites
- L18:251 excites
- L23:1936 suppresses
- L23:126 suppresses
- L19:400 excites
- L8:328 excites
- L13:2366 suppresses
- L10:669 suppresses
- L18:1292 excites
- L21:2617 suppresses
- L9:74 excites
Feeds later layers
- L27:2277 strongly suppresses
- L27:1606 suppresses
- L27:1050 suppresses
- L27:2506 suppresses
- L26:408 weakly suppresses
- L26:440 weakly suppresses
- L26:2147 weakly suppresses
- L27:2055 weakly excites
- L27:1229 weakly excites
- L26:706 weakly excites
- L26:1474 weakly suppresses
- L27:1917 weakly excites
- L27:597 weakly excites
- L27:1488 weakly suppresses
- L26:2671 weakly excites
- L26:1369 weakly excites 'Pr' starting capitalised names
- L26:2641 weakly suppresses
- L26:2255 weakly suppresses
- L26:506 weakly excites
- L26:2998 weakly excites
- L27:908 weakly excites
- L27:3034 weakly excites
- L26:246 weakly suppresses
- L27:2559 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
- edException
- etail
- ritis
- passwd
- ByExample
- ponge
- onSelect
- unky
- ting
- igit
Suppresses
- 来
- 來
- 去
- 来做
- come
- comes
- 来的
- 来找
- coming
- 来进行