sentence-initial 'Not' starting a clause
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. 23 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 the word 'Not' when it begins a new sentence or clause, as in 'Not surprisingly' or 'Not too long ago'. When it was turned up, the model started repeating the word 'sure' over and over.
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.
-
20th / 21st-century philosophy|⏎|Main interests||Linguistics · Psychology⏎Philosophy of language⏎Politics · Ethics⏎|Notable ideas||Generative grammar|⏎Avram Noam Chomsky (born
-
of limiting individual impact on climate change.⏎The event hosted three food sustainability specialists Julia Davies (Nottingham Trent University), who gave a global food security perspective, Hannah Ford (University
-
try and eat more produce. Of course, fruits and vegetables supply vitamins, minerals, and fiber, which we need to achieve and maintain overall health, too.⏎Not surprisingly, a steady intake of fruits
-
associated with investments in renewable energy. And in countries with a low GDP per capita, nuclear electricity production clearly associates with CO2 emissions that tend to be higher.”⏎Not surprisingly, the study attracted criticism. Some summarily dismissed it as pseudos
-
States and Mexico; The opposite occurs with Cubans and Puerto Ricans living in the United States.⏎This is reflected in a survey carried out between Axios and Noticias Telemundo. Cubans have settled very well in Florida since the times
-
2011, the UN Special Representative to Kosovo called for an independent investigation into Kosovan organ trafficking allegations to be spearheaded by the UN Security Council.⏎Noting the changing forms of slavery, including the rise in global sex trafficking, UN
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: I can still see some former students in the former school, and I also can feel some feelings of stillness and separation between former and new, or maybe the former and latter parts of the former school. I have both feelings and feelings of
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
to protect the earth’s climate. The global climate crisis is non-negotiable. #Move the Date: We need to do everything we can
-
the damage.⏎Emergency preparation is now an important part of a city's defense. We can no longer depend on historical weather patterns to be a guide. Sea-level rise and climate
-
: Sometimes treats⏎There are certain foods that your bunny should only have in moderation such as foods that<|endoftext|>Moving the boundaries (on Shoftim)⏎"You shall not move the boundary of your neighbour, which the first ones made his boundary when you
-
soliloquy in Act 1 Scene 2 where he recognizes that "self-slaughter" is against God's commandment (i.e. Thou shall not kill). Here, Hamlet is being over dramatic: youthful hyperbole. His
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
- L20:1414 excites
- L22:635 suppresses
- L4:151 excites
- L15:786 excites
- L17:443 excites
- L4:9 excites
- L11:1095 suppresses
- L16:752 excites
- L10:111 suppresses
- L10:504 suppresses
- L10:386 excites
- L17:1644 suppresses
- L1:509 excites 'to' after 'unable'
- L19:259 excites
- L18:2220 suppresses
- L9:54 excites
- L18:1763 excites comma in 'not X, but Y' contrasts
- L19:2184 suppresses
- L14:2441 excites
- L20:212 weakly excites
- L21:2799 weakly suppresses
- L7:32 weakly excites
- L0:2681 weakly excites
- L15:1293 weakly suppresses
Feeds later layers
- L25:1756 weakly excites
- L25:2845 weakly excites
- L25:1264 weakly suppresses
- L25:1985 weakly excites
- L27:1117 weakly excites
- L27:618 weakly excites
- L27:2268 weakly excites
- L26:670 weakly excites
- L25:1094 weakly excites
- L26:2916 weakly excites
- L25:1096 weakly suppresses
- L24:590 weakly suppresses
- L27:2044 weakly excites
- L26:677 weakly excites
- L27:938 weakly excites
- L24:960 weakly excites
- L27:2827 weakly excites
- L26:1395 weakly suppresses
- L24:2690 weakly excites
- L26:3012 weakly excites
- L24:1964 weakly excites
- L25:1674 weakly excites
- L26:1975 weakly excites
- L27:398 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
- sure
- sure
- Sure
- 太
- Sure
- ched
- ches
- 那么
- CHED
- 太高
Suppresses
- 前者
- ◄
- べき
- offee
- ainer
- 后者
- plier
- rowsing
- ⊘
- former