capital 'K' starting a name piece
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
This neuron responds to a capital K that begins a name or word piece, such as the K in Kaliangara or Koguryo. Its steering results do not show a clear theme, so what it does for the text 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.
-
presentation at the conference, Jeff Schinske, professor of biology at De Anza Community College, and three current or former students — Amanda Snyder, Jahana Kaliangara, and Monica Cardenas — described an intervention designed to enhance students
-
culture towards Japan. In the year 660 it was conquered by Silla 6, which in 668, with the capitulation of Koguryŏ, was able, for the first time in history, to unify
-
Science Identity in a Diverse, Community College Context⏎Jeff Schinske, Amanda Snyder, Heather Perkins, Mary Wyner, Monica Cardenas, Jahana Kaliangara—all of De Anza College⏎De Anza Community College,
-
|View the image Gallery|⏎Flowing out of the mountains surrounding the Native community of Kake, Gunnuk Creek meanders through this seaside town and on into Keku Strait. The creek provides fresh water for the municipality and the water supply
-
(Bhavanasunya). Manu starts getting involved in various emotions, thoughts and actions. These are sequentially portrayed with Shraddha, Ida, Kilaat and other characters playing their part, contributing in them. The chapters are
-
associated with multiple sclerosis⏎Science | January 13, 2022⏎Bjornevik K, Cortese M, Healy BC, Kuhle J, Mina MJ, Leng Y, Elledge SJ,
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: Even while naturally good naturally good good good good good good good good good good good good good good good good good good good good good good good good good good good good good good good good good good good good good good good good good good good
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
of events.”⏎The new normal⏎“This analysis confirms what we already know – that we are suffering an extreme weather event caused by climate change,” says Gareth Redmond-King, a climate change specialist at WWF. “
-
only. Though<|endoftext|>نوع مقاله : مقاله پژوهشی⏎1 دانشیار معماری، دانشکده معماری و شهرسازی، دانشگاه علم
-
.⏎2 پژوهشگر دوره تخصصی دکتری معماری منظر، دانشکده معماری و شهرسازی، دانشگاه علم
-
children vaccinated against COVID-19.⏎In an opinion piece for The Hill, Senator Roger Marshall (R-Kan.) says COVID-19 immune response research is important, as well as
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
- L22:2170 strongly excites
- L21:590 excites
- L18:816 suppresses
- L20:613 suppresses
- L26:495 suppresses
- L19:381 excites word-start 'acc' fragments
- L20:265 weakly excites
- L25:2286 weakly excites
- L18:569 weakly excites
- L22:56 weakly suppresses comma after a clause opener
- L20:1120 weakly excites
- L20:582 weakly excites
- L4:1860 weakly excites
- L20:61 weakly suppresses
- L16:7 weakly excites
- L20:2297 weakly excites
- L0:2999 weakly suppresses
- L10:1795 weakly suppresses
- L0:2385 weakly suppresses
- L16:61 weakly excites
- L17:1216 weakly suppresses
- L0:1996 weakly excites
- L19:2402 weakly excites
- L16:74 weakly suppresses
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
- 很高
- 感染者
- .retrieve
- terior
- comed
- 素
- ieval
- ::$
- 很差
- 很低
Suppresses
- umar
- огда
- han
- おく
- mor
- irk
- 本身就
- aiser
- -priced
- cing