'://' in web addresses
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. 21 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 colon and slashes that come right after 'http' or 'https' in a web address. Turning it up or down did not make the model write about links or web addresses, so the steering output does not confirm the pattern.
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.
-
1: Download Drupal⏎First download the Drupal using following command:⏎# wget http://ftp.drupal.org/files/projects/drupal-7.33.zip⏎Step
-
413967-1-768x146-1-300x57.png 0 0 admin https://
-
://my.fsf.org/donate/working-together/octave.⏎DocumentationUser manual included; User FAQ included; Printed and online user manual available from http://www.network-theory.co.uk/octave/manual/⏎- PARI GP⏎-
-
to supports all kinds of powerful, dynamic applications and services. For more information on this great topic, please consult:⏎- Sun's Java Server Pages home (http://java.sun.com/products/jsp/)⏎- Westy Rockwell's article on "
-
structure = function.⏎(Image from: https://commons.wikimedia.org/w/index.php?curid=12679
-
restriction of intentional use in many industrial sectors as detergents, fertilizers, construction and oil&gas materials. The documents are available at this link (https://echa.europa.eu/it/registry-of-restriction-intentions/-/dis
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.m. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
there, I get an exception, which is life as usual for a Python programmer.⏎if not server_response['successful']:⏎etag = server_response['file_handle'].info().getheader('ETag')⏎xml_document = server_response['xml_document']
-
achieved with the following code:⏎TYPE List = REF RECORD next: List; value: INTEGER; END;⏎VAR x := NEW(List, value:=1); BEGIN x.next :=
-
syntax in configuration files is case-sensitive.⏎The machine configuration file, Machine.config, contains settings that apply to an entire computer. This file is located in the %runtime
-
, instead of writing Users.Current.User.Name, you can just write Users.Current.Name.⏎I’ll grant you that this doesn’t seem much for the User entity
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
- L16:1696 excites
- L18:2709 suppresses
- L19:1049 suppresses
- L20:2113 excites
- L17:2932 excites
- L19:913 excites
- L19:1766 suppresses
- L12:1039 excites
- L16:2648 suppresses
- L19:1400 excites
- L20:1340 suppresses
- L11:2255 excites
- L19:3009 excites
- L14:1248 suppresses
- L9:2559 excites
- L19:2234 excites the :// in web addresses
- L15:2615 excites
- L11:1047 weakly excites pieces of software and library names
- L19:274 weakly suppresses
- L8:2364 weakly excites
- L10:845 weakly excites
- L18:1565 weakly suppresses
- L15:653 weakly excites
- L14:2515 weakly suppresses
Feeds later layers
- L24:228 strongly suppresses
- L26:593 strongly suppresses
- L25:294 suppresses
- L23:372 suppresses
- L27:362 suppresses
- L26:174 suppresses
- L27:580 suppresses
- L26:943 suppresses
- L26:1169 suppresses
- L27:1328 suppresses
- L25:387 suppresses
- L27:816 suppresses
- L25:412 suppresses
- L27:351 suppresses
- L27:875 suppresses
- L26:316 suppresses
- L25:1135 suppresses
- L27:370 suppresses
- L27:1240 suppresses
- L25:656 suppresses
- L27:2971 weakly suppresses
- L22:1829 weakly suppresses
- L26:167 weakly suppresses
- L26:653 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
- わり
- 倒在
- ностей
- 同一个
- Meadows
- 人々
- 켈
- 許多
- 否則
- 好几个
Suppresses
- .s
- .S
- .b
- .Mon
- .F
- .f
- .L
- .M
- .m
- .v