"Join" starting a call to action
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. 5 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them 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 the word "Join" when it starts a sentence inviting the reader to take part in something, like "Join us as we explore...". When it was turned up, the writing drifted into unrelated topics and ran on without much sense.
Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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.
-
redhorse — long known by the Cherokee and only recently recognized by western science — is an iconic species of the Little Tennessee watershed. Join Luke Etchison of the NC Wildlife Resources Commission to learn about science and restoration
-
, use the search bar, or filter options on the left side of the page.⏎Are you looking for a unique way to teach storytelling elements to your students? Join us for a literary coding adventure as we explore how to use Ozobot to learn
-
us to various diseases and ailments. From cancer to seizures, these remarkable animals can detect subtle changes in our scent or behavior that can indicate an underlying medical issue. Join us as we delve into the world of medical detection dogs
-
world, for as long as you wish!⏎Pinyon Jay Project *⏎Community scientists play a critical role in gathering crucial data on a scale not otherwise possible! Join Audubon in-person or virtually for an orientation to Pinyon Jays and
-
actions.⏎Edpuzzle gives you the opportunity to easily create interactive video lessons for your students that you can integrate right into your Learning Management System (LMS). Join us as we explore how to use Edpuzzle’s built-in question feature to
-
In this article, we will explore the incredible breathe-holding ability of turtles and discover how they sleep and eat underwater effortlessly. Join us on this underwater journey as we unveil the secrets behind how long can turtles breathe underwater.⏎How
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: when I get 32.2215, 3.2215, 3.2215, 3.2215 and the rest of the... I have... 2.0
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
"What would you do if you were stuck in one place, and every day was exactly the same, and nothing that you did mattered?" Phil asks two local guys at the<|endoftext|>Courtesy of EarthSky⏎A Clear Voice for Science⏎Here is
-
years. Teaching fills her life and gives her a sense of accomplishment, but recently she has begun to forget details and has become more and more disorganized. At first, she laughed it off, but her memory problems have worsened. Her family
-
by Susanne Bacon.⏎As a kid and young teenager, I loved playing soccer. The boys in my class even let me play on their teams as the only girl during our
-
the South. The fight for civil rights would never be the same.⏎As the sit-ins grew in strength, the department stores began to change their tactics. At first, the stores had responded to the protesters by ignoring them or closing down the lunch
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
- L1:2572 weakly suppresses
- L4:83 weakly excites
- L2:90 weakly excites
- L4:229 weakly excites
- L2:338 weakly excites
- L3:23 weakly excites
- L0:2915 weakly suppresses
- L1:1395 weakly suppresses
- L4:21 weakly excites
- L4:2021 weakly suppresses
- L4:660 weakly suppresses
- L3:2668 weakly excites
- L0:2656 weakly excites
- L1:1505 weakly suppresses
- L3:2599 weakly excites
- L0:1054 weakly excites
- L0:359 weakly suppresses hyphen inside COVID-19 and similar compounds
- L3:424 weakly suppresses
- L2:641 weakly suppresses
- L1:3045 weakly excites
- L4:29 weakly suppresses
- L0:3065 weakly suppresses
- L0:2572 weakly excites
- L3:294 weakly excites
Feeds later layers
- L7:1198 excites
- L9:2945 excites
- L17:1139 weakly suppresses
- L19:1131 weakly excites
- L8:1139 weakly excites
- L6:570 weakly excites
- L7:160 weakly suppresses
- L14:724 weakly suppresses
- L19:2604 weakly suppresses
- L15:3009 weakly suppresses
- L24:808 weakly excites
- L21:3018 weakly suppresses
- L10:2002 weakly excites
- L12:1140 weakly excites
- L11:1582 weakly excites
- L9:512 weakly suppresses
- L20:2160 weakly suppresses
- L7:2880 weakly excites
- L18:3049 weakly excites
- L17:980 weakly excites
- L12:1224 weakly suppresses
- L10:3050 weakly excites
- L10:1010 weakly excites 'in the process of' and continuing actions
- L13:2291 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
- 妫
- reign
- sdale
- secs
- realms
- fin
- sons
- (chars
- esen
- gears
Suppresses
- ист
- ⼝
- 吸毒
- 引用日期
- 北约
- istica
- 对我说
- 跡
- tendency
- 心脏病