unclear: spaces in mixed-language text
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. 27 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
The pattern is hard to pin down. The strongest moments come in messy multilingual pages, such as dictionary entries, glossaries, and web boilerplate, at a point where a new chunk of text starts. When turned down, the model wrote long runs of blank lines, so the neuron seems tied to whitespace and line breaks in some way.
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.
-
�� : sleep 不 : not 觉 : to feel, wake up 晓 : dawn⏎Meaning for Lien 2 处 : place 闻 : hear 啼 : cry 茑 : bird⏎Meaning for
-
oo⏎DATE ADDED: Jan 20, 2011⏎TRANSLATED BY: まだ⏎J-GO EDITED BY: まだ⏎If you're going to give this activity⏎a low-rating, please
-
Bubbling fluidized bed boilers; Circulating fluidized bed boilers⏎- Front<|endoftext|>CNY - 中國元,人民幣⏎最常用 CNY 交叉匯率⏎為什麼您
-
and Silla 新羅 5, was the most faithful ally of Japan in the peninsula for fighting against the
-
is to click here and submit your correction.⏎Frequently Asked Questions (FAQ)⏎What do you mean by communized?⏎Meaning of communized is بات چیت کی⏎Whats the definition of communized?⏎Definition of the
-
– that they might know.⏎4 tn Heb “and they did so.”<|endoftext|>Communized meanings in Urdu⏎Communized meanings in Urdu is بات چیت کی Communized in Urdu. More meanings of communized,
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.
-
commission, Grade was entrusted with the remains of the dismantled schooner Wawona.⏎Built by influential shipbuilder Hans Ditlev Bendixsen, the 165 ft long schooner Wawona sailed from 189
-
, reports, other regulations, websites or to contact your legislators.⏎ADA Anniversary Tool Kit includes: Overview of the Americans with Disabilities Act and the ADA Amendments Act of 2008; ADA Findings, Purpose, and History; The ADA
-
oper Tire test center⏎It became particularly strong during World War II--when advancing Japanese forces cut off U.S. access to hevea--and during the 1970s.⏎Guayule grows much faster than hevea
-
ves, especially in America, where women like Sally Jackson, Mary Keehn, and Laura Chenel launched the "American Artisan" cheese movement back in the 70's.⏎This month, we honor women makers from all over the world
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
- L22:824 strongly excites blank before an expected number
- L19:581 strongly suppresses blank slot before a number
- L22:372 strongly excites empty spot before a number or "3D"
- L23:77 strongly excites the space right before a number
- L18:377 excites number right after 'about' or 'less than'
- L20:344 excites hyphen inside compound words
- L26:97 suppresses
- L21:339 excites spot just before a year
- L25:33 excites spot where a number is due
- L17:285 suppresses
- L18:106 suppresses
- L22:90 suppresses
- L4:53 excites blank slot before a number
- L25:222 excites
- L24:248 excites
- L23:109 excites
- L22:216 excites
- L22:265 excites
- L8:111 suppresses the space before a year in dates
- L25:265 suppresses
- L21:321 suppresses
- L19:697 suppresses
- L17:402 suppresses
- L18:2403 excites
Feeds later layers
None: this is the last layer.
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
- �
- 通讯员
- #
- )){⏎
- 皑
- ernels
- )))⏎
- orted
- ernel
- urved
Suppresses
- ⏎
- ⏎
- ⏎
- ⏎⏎
- ⏎
- ⏎
- ⏎
- ⏎⏎