hyphen joining two words
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. 2 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 lights up on a dash that joins two words into one compound, like 'drama-in-education' or 'match-the-phrase'. Turning it up made the model's text drift without any clear theme.
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.
-
efforts of diverse Polish groups -Catholics, socialists, independents and others- to save Jewish lives despite great personal risks.⏎The London-based Polish Government-in-
-
Pathways: Children, Pedagogy and Play in the Age of Digital Technology⏎Keywords:pedagogy, children, play, day camp, drama-in-education⏎Children today are immersed in a digital environment that is marginalizing and,
-
to a glass of water, representing "tengo sed," or "I am thirsty."⏎The app also challenges its students to answer increasingly difficult typing, match-the-phrase-to-the-photo, listening, and conversational tests.⏎"With CatAcad
-
Extract Compressed Drupal⏎Extract the Drupal zip file in the /var/www/html directory using following command:⏎# cd /var/www/html⏎# unzip drupal-7.33.zip⏎It will create a folder as /var/www/html
-
chemical,petroleum,textile,and foodindustry sources.Reference mststeel/304-stainless-steel-properties/ results for this questionFeedback304 Stainless Steel Plate Midwest Steel
-
to show how the levels translate to real learning environments.⏎Through the long-term athlete development model, the Player Development Curriculum addresses the topic of proper practice/training-to-competition ratios. USA Basketball has defined competition as the act of competing against another team
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: it. In China, the market is our customers, are the main. So the problem. So the problem. So the problem. So the problem. So the problem. So the problem. So the problem. So the problem. So
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
Labor, The Economics Daily, Demographics of Gulf War-era II veterans at https://www.bls.gov/opub/ted/2010/ted_20100805.htm (visited November 27,
-
, houses of worship, and religious schools. The full text of RLUIPA is available at http://www.usdoj.gov/crt/housing/housing_rluipa.htm. Below is a summary of the law’s key provisions, with
-
Wedlock,” Christina Gibson-Davis. Demography, July 2, 2014. DOI: 10.1007/s13524-014-0308-7<|endoftext|>
-
31(1), 131-132.⏎Available at: https://scholarworks.uni.edu/pias/vol31/iss1/25<|endoftext|>New images from NASA’s Mars Reconnaissance Orbiter show finger
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
- L0:1301 suppresses
- L0:178 suppresses
- L0:1224 suppresses
- L1:2683 weakly excites
- L0:0 weakly excites
- L1:1096 weakly suppresses
- L1:2927 weakly suppresses
- L1:2568 weakly excites
- L1:576 weakly excites
- L1:146 weakly excites
- L1:824 weakly suppresses
- L0:1518 weakly excites
- L1:1699 weakly excites
- L1:2099 weakly excites
- L1:331 weakly suppresses
- L1:761 weakly suppresses
- L1:1355 weakly excites
- L0:46 weakly excites
- L1:1591 weakly suppresses
- L1:1492 weakly excites
- L1:2244 weakly suppresses
- L1:2940 weakly suppresses
- L1:1387 weakly suppresses
- L1:2801 weakly excites
Feeds later layers
- L11:102 excites
- L10:134 excites
- L4:58 excites
- L26:186 excites
- L9:11 weakly excites
- L3:195 weakly suppresses
- L4:0 weakly excites
- L17:927 weakly excites
- L3:48 weakly excites
- L17:364 weakly excites hyphen joining two words
- L25:1710 weakly suppresses
- L24:2894 weakly suppresses
- L16:975 weakly excites
- L6:1262 weakly excites
- L6:590 weakly excites
- L27:691 weakly suppresses
- L4:46 weakly excites
- L23:1418 weakly suppresses
- L6:1466 weakly excites
- L3:254 weakly excites
- L9:31 weakly excites
- L8:687 weakly excites
- L4:6 weakly excites
- L23:2325 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
- ">@
- init
- ropa
- "><!--
- 操场
- 惯例
- ">⏎⏎⏎
- 师傅
- 誓言
- 无辜
Suppresses
- bose
- shire
- tele
- atori
- diffuse
- 雳
- borne
- ImageSharp
- multid
- ington