the :// 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. 19 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 reacts to the colon and slashes in the middle of a web address, right after words like http or https. Turning it up did not produce anything web-related in the sample outputs, so its effect on writing is unclear.
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.
-
[change | change source]⏎- "Chomsky is citation champ". MIT News Office. 1992-04-15. http://web.mit.edu/newsoffice/1992/citation-0
-
1: Download Drupal⏎First download the Drupal using following command:⏎# wget http://ftp.drupal.org/files/projects/drupal-7.33.zip⏎Step
-
is available for this purpose (e.g., US Census data at https://www.census.gov/) and can be of tremendous help in targeting various ethnic groups
-
structure = function.⏎(Image from: https://commons.wikimedia.org/w/index.php?curid=12679
-
for efficient DNA cloning and recombineering in E. coli". PLoS ONE 8 (2): e57075. . . http://www.plosone.org/article/info%3Adoi%2F10
-
lenses, similar to radiotelescopes arrays.⏎The limitation is givven by the Index of Refraction. For more information you can explore this: https://en.wikipedia.org/wiki/Refractive_index⏎Regarding the image, it is
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: how to actually * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
, rule, minor head injury⏎Ann Afr Surg. 2018;15(2):57-61 DOI:http://dx.doi.org/10.4314/aas.v15i2
-
Red List of Threatened Species 2016: e.T6794A21952571. https://dx.doi.org/10.2305/IUCN.UK.2
-
/2002, WGBH Media Library & Archives, accessed March 29, 2017, http://openvault.wgbh.org/catalog/V_4E4EE3EAD24043
-
Recommendation on<|endoftext|>- OurWorld 2.0 - http://ourworld.unu.edu/en -⏎What does Cancún offer the climate generation?⏎Posted By Huw
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
- L10:2362 suppresses
- L10:814 suppresses
- L5:2579 excites
- L15:1962 suppresses
- L18:2292 suppresses
- L10:613 excites
- L18:569 excites
- L6:2232 weakly excites
- L0:1536 weakly excites
- L15:2745 weakly excites
- L14:2306 weakly suppresses word after a motion or location verb
- L4:26 weakly excites
- L5:50 weakly excites
- L11:2017 weakly excites
- L3:456 weakly suppresses
- L9:1040 weakly excites
- L9:656 weakly suppresses
- L6:18 weakly excites
- L14:1505 weakly excites
- L18:1451 weakly excites
- L4:2427 weakly suppresses
- L13:1372 weakly excites number before an age or week count
- L4:1796 weakly suppresses
- L5:120 weakly suppresses
Feeds later layers
- L24:228 excites
- L21:1855 excites
- L20:2623 excites
- L20:536 excites
- L21:1033 excites '://' in web addresses
- L25:656 excites
- L22:406 weakly excites
- L23:372 weakly excites
- L26:1066 weakly excites
- L27:580 weakly excites
- L25:294 weakly excites
- L27:362 weakly excites
- L20:1111 weakly suppresses
- L26:593 weakly excites
- L23:1308 weakly excites
- L27:879 weakly excites
- L21:389 weakly excites
- L26:316 weakly excites
- L24:1427 weakly excites
- L25:1135 weakly excites
- L20:327 weakly excites
- L23:2581 weakly excites
- L20:2748 weakly suppresses
- L21:1825 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
- www
- www
- .example
- �
- ền
- -static
- .stack
- ritis
- .bias
- example
Suppresses
- ertino
- 还是要
- 其实
- 依旧
- 其實
- 依然
- 属实
- 实际
- LEncoder
- 仍