colon after a metadata label
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. 9 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 that follows a label in a document's details, such as "Received:", "Published:", "Date:" or "Created on:". Turning it up made the model's writing wander into unrelated text, so the steering does not show this 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.
-
the Pharaohs⏎World bibliography<|endoftext|>Also found in: Dictionary, Thesaurus, Medical, Wikipedia.⏎cavities connected to the
-
930s.⏎Images: (Left) via the National Media Museum; (Right) via Getty Images<|endoftext|>
-
0.3390/ijerph9114170⏎Published: 15 November 2012⏎Abstract: Understanding the effects of
-
.”<|endoftext|>Rural earthen roads impact assessment in Phewa watershed, Western region, Nepal⏎© The Author(s). 2016⏎Received: 23 November 2015⏎Accepted: 26 May
-
*Corresponding Author:⏎- Yeongha Wook⏎Department of Pediatrics,⏎Soroka University Medical Center,⏎E-mail: [email protected]⏎Received: 01-Nov-2022, Manuscript No. AAAJ
-
is determined" (Zuck, pp. 20). F. Biblical Theology "Biblical theology is the study of divine revelation as File Size: KB.⏎Get this from a library! Principles of biblical interpretation: (sac
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: so I want to know if you can say that you are interested in reading about the latest trends in the technology industry industry industry technology industry industry industry industries industries industries industries industries industries industries industries industries industries industries industries industries industries industries industries industries industries industries
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
them at risk of tearing suture lines, bleeding or indeed, in extreme circumstances perforation. Stretching intestinal or gastric mucosa may also result in haemorrhage.⏎To avoid such complications, travel should be avoided for 10 days
-
the gastrointestinal system. Their results are published in "Nature Structural & Molecular Biology".⏎"These findings could in principle be explored to develop a decoy alternative to conventional
-
,10 250,190 160,210" style="fill:lime;stroke:yellow;stroke-width:1"/> </svg>⏎Here is what the above code looks like in HTML5
-
Armstrong said.⏎The complex research appears in the journal Nature Physics.<|endoftext|>Yoga is a form of exercise that focuses on strength, flexibility and
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
- L4:193 strongly excites
- L3:77 excites
- L5:134 suppresses
- L8:16 suppresses
- L0:2941 weakly excites
- L2:872 weakly suppresses
- L4:946 weakly suppresses
- L0:2833 weakly suppresses
- L8:1076 weakly suppresses
- L6:2596 weakly suppresses
- L6:1104 weakly excites
- L1:2827 weakly suppresses
- L8:1627 weakly suppresses
- L3:841 weakly suppresses
- L5:454 weakly suppresses
- L4:688 weakly excites "to" after begin or start
- L5:2139 weakly excites
- L6:2277 weakly suppresses
- L1:1430 weakly suppresses
- L6:2773 weakly excites
- L6:12 weakly excites
- L4:2137 weakly excites
- L4:2148 weakly excites
- L6:65 weakly excites
Feeds later layers
- L17:338 excites colon introducing an explanation
- L10:206 excites colon or equals sign before a definition
- L18:2254 excites colon before a subtitle
- L14:1544 excites
- L11:1307 excites 'words' in 'in other words'
- L16:1209 excites
- L16:1608 suppresses
- L20:1772 excites
- L20:2332 excites
- L17:806 excites
- L17:2613 excites
- L12:1506 excites
- L19:1790 excites
- L16:710 excites
- L14:409 suppresses
- L15:261 excites
- L10:2009 excites
- L14:942 excites
- L15:1299 excites
- L21:336 weakly excites
- L19:1755 weakly excites
- L11:1072 weakly excites
- L11:502 weakly excites
- L11:1787 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
- eselect
- Sense
- elihood
- emin
- 钉
- Align
- *\
- Administration
- cap
- :\/\/
Suppresses
- oupon
- rowable
- ApiOperation
- upd
- _hpp
- roup
- iffies
- 等奖
- InSection
- ó