'just' meaning barely or recently
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. 20 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 word 'just' when it means 'barely', 'only a moment ago' or 'recently', as in 'could just make out' or 'just last month'. Turning it up did not produce any clear change in 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.
-
my eyes in the darkness, I struggled to find the source of the commotion. Through my binoculars, I could just a make out the silhouette of a subadult grizzly devouring salmon carcasses
-
Do The Right Thing” was set in the same multicultural Brooklyn as “The Cosby Show,” but now it was a simmering cauldron of racial tension lying just beneath the surface. The movie introduced us to another key element of Critical Race Theory
-
transport.”⏎Private firms have been working on commercial space travel for a while, with the SpaceX Dragon capsule successfully completing its first resupply mission to the International Space Station just last month, but recently there’s been
-
with the non-emotional cognitive task.⏎"One of the reasons why these results are important is that we know that young people are getting much less sleep than they did just ten years ago," explains Dr Petrovic. "<|endoftext|>This October, celebrate Global
-
ocious.⏎Perhaps it is time to stop reading my words, bundle up the kids and hike into the Arctic. Why not today! Children will almost certainly see seals just yards away and, if the timing is right, go nose to nose with a
-
picturesque backdrop for pioneering comic strip Krazy Kat and for many a Windows Vista desktop, in photos of everyone from the Marlboro Man to a family of smiling tourists just yesterday.⏎The valley is found in the heart of the Navajo Nation, and
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: the more the MCMD of a set of numbers, the smaller the numbers are. Then, the larger the MCMMD, the larger the numbers are. But that is a very general statement, but the ones on the subject of
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
uously displayed, rode on toward the gray-clad horsemen.⏎“I’ll wager that’s General Lee,” said the captain, with a glance at me. “Let us hope things turn out all right.”⏎As the two Federals neared the
-
to death. But, instead of putting him to death, Ruben, the oldest brother, hoping to come back later and rescue Joseph says, "No, letís not put him to death - letís throw him into a pit." And
-
the body, and consequently a bend is formed at the<|endoftext|>DENVER – Put on your poodle skirts and tune in Elvis on the transistor radio, because it’s
-
- Healthy Happy Me: Easy-Peasy guide to awesome health Regular Price 150.00 Special Price 135.00⏎- Let's Get Active : Learning about Flowers (An illustrated activity book that teaches young learners
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
- L10:1610 excites
- L15:484 excites
- L10:1278 weakly excites
- L15:1693 weakly excites
- L10:556 weakly suppresses
- L15:1627 weakly suppresses
- L1:1583 weakly excites
- L18:2158 weakly suppresses
- L9:1609 weakly suppresses
- L12:78 weakly excites
- L19:145 weakly excites
- L16:2860 weakly suppresses
- L0:2941 weakly excites
- L13:446 weakly suppresses
- L4:236 weakly suppresses
- L19:992 weakly suppresses
- L6:907 weakly excites
- L9:1293 weakly excites
- L9:915 weakly excites
- L4:398 weakly excites
- L6:19 weakly suppresses second half of proper names
- L10:678 weakly suppresses
- L18:220 weakly excites
- L16:1011 weakly suppresses
Feeds later layers
- L22:2495 weakly excites
- L24:918 weakly excites
- L23:2802 weakly suppresses
- L25:1588 weakly suppresses
- L27:2018 weakly suppresses
- L22:2314 weakly suppresses
- L23:1645 weakly suppresses
- L27:1213 weakly excites
- L22:2299 weakly excites
- L23:392 weakly suppresses
- L23:345 weakly excites
- L22:3024 weakly excites
- L25:3044 weakly excites
- L22:720 weakly suppresses
- L23:1129 weakly suppresses
- L24:952 weakly excites
- L25:218 weakly excites
- L25:1425 weakly suppresses
- L21:2180 weakly suppresses
- L21:1295 weakly suppresses
- L23:1936 weakly suppresses
- L21:875 weakly suppresses
- L21:2051 weakly suppresses
- L25:1373 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
- 现在
- possibly
- horia
- ário
- 接下来
- ','',
- →ar
- getNext
- ランス
- maybe
Suppresses
- THAN
- THEN
- THEN
- Than
- than
- those
- than
- ucid
- -than
- 則