line break after a title
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. 0 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 reacts to the line break right after a short title or heading, just before the body text starts. Turning it up made the model's writing drift into unrelated topics, with no clear link to the trigger.
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.
-
essayist,<|endoftext|>How to Prevent Dry Eyes During Summer Months⏎Summer is around the corner, and one can feel the heat of the scorch
-
. These foods include dairy products, legumes and grains.⏎How Much Protein On Paleo Diet⏎The goal of the paleo diet is to return to a way of eating that
-
<|endoftext|>Genetic Testing Resources⏎Inherited retinal degenerative diseases, like retinitis pigmentosa or St
-
Manipulation⏎- Spirit Physiology⏎- Undead Physiology⏎- Despite being able to survive without
-
Initiative Prevention Curriculum⏎Target Audience: Middle school-aged youth⏎Administration: This curriculum can be used by
-
ale Revolution and Climate Change⏎- The Shale Revolution and Earthquakes⏎- The Economic Impacts of
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 will not be able to do that, I will not be able to do that. I will not be able to do that. I will not be able to do that. I will not be able to do that. I will not
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
inium which is fairly easy to identify and is blooming now.⏎Deerberry – Vaccinium stamineum – has relatively large, attractive, abundant, white bell-shaped flowers
-
that identification is difficult and natural hybrids are common. Further, huckleberries, which are now in the genus Gaylussacia were once in the genus Vaccinium. And one of the Vacciniums I think I may have identified here goes
-
").⏎In 1608 tried to patent a telescope fitted with a pair of sunglasses, a concave and another convex, which it termed perspicillum: was born the first astronomical telescope refractor. The manufacturer of lenses designed it with
-
're saving.⏎For More Information...⏎Phone: 1-888-STAR-YES (1-8
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
None: this is the first layer.
Feeds later layers
- L1:272 suppresses
- L14:387 suppresses
- L22:2375 suppresses
- L18:1370 weakly excites
- L2:2713 weakly suppresses
- L10:783 weakly suppresses
- L16:84 weakly excites
- L14:2319 weakly excites
- L16:2546 weakly excites
- L1:655 weakly suppresses
- L24:1117 weakly excites
- L14:2029 weakly excites
- L3:2516 weakly suppresses
- L14:371 weakly suppresses
- L2:368 weakly suppresses
- L8:2337 weakly suppresses
- L24:1084 weakly suppresses
- L2:2852 weakly suppresses
- L13:753 weakly suppresses
- L1:918 weakly suppresses
- L27:2560 weakly suppresses
- L13:152 weakly excites
- L1:138 weakly suppresses
- L2:1102 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
- mıştı
- ctors
- �
- 十条
- erner
- izedName
- 韬
- estad
- ertools
- 中华文化
Suppresses
- лина
- 捡
- imit
- 凇
- 彗
- Sprites
- }:
- cones
- avit
- ints