the "un" prefix splitting a word
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. 26 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 letters "un" when they start a word stem that continues after them, like "unearned" or "unspoken" split off from the rest of the word. Turned up, the model got stuck repeating fragments like "ly" and "out of".
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.
-
with 100% plagiarism free content. We prepare quality content and notes for Characteristics of Linear Programming topic under Operations Management theory and study material. These are avail for subscribed users and
-
anything but an asset. Cash is a current asset, and serves to increase the net worth of whoever is in possession of the cash. But when the cash is unearned, or you haven't performed the service or provided the product that the cash
-
this new field are also valid. Some community nervousness is certainly justified since “these technologies offer the potential for great benefit, [but] it is not benefit unalloyed.”12⏎However, banning research is also unlikely to be
-
(Mt. 12: 41).<|endoftext|>“There shall never be any bond slaverie, villinage or Captivitie amongst us unles it be lawfull Captives taken in just warres, and such strangers
-
as an energy source. Researchers had believed that keeping the donor and acceptor layers as pure as possible was the best way to ensure that the excitons could travel unimpeded, so that solar cells could capture the maximum amount of energy.⏎NC
-
Too Much”⏎Some people feel more than others.⏎They are often told—spoken and unspoken—that they are ‘too much,’ ‘too intense,’ ‘too sensitive,’
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: what can be in outfit同用同用同用同用同用同用同用同用同用同用同用同用同用同用同用同用同用同用同用同用同用
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
the food we consume which regularly introduces new elements into this already-unstable milieu.⏎Disruption to gut flora is known to have an adverse effect on
-
windows provide natural daylight, and the sun's rays help keep rooms warmer in winter. If you want more light, add insulated windows to the home's south side-unless you live in the desert or an extremely hot
-
-only models.⏎But why would any school board adopt such an obviously flawed sex education program?⏎For one, American society has an extensive history of supporting abstinence-only-until-marriage (AOUM) programming. This is the result of
-
, breathing exercises, pranayama (breath control), and self-control. When you have almost any queries concerning wherever along
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
- L16:3 suppresses
- L2:761 suppresses
- L20:61 suppresses
- L20:212 excites
- L4:216 excites
- L20:1414 excites
- L16:160 excites
- L10:431 excites
- L4:151 excites
- L18:853 suppresses
- L19:118 excites
- L17:391 excites
- L18:1104 suppresses
- L1:259 suppresses
- L21:625 excites
- L0:1168 suppresses
- L22:271 suppresses
- L5:499 excites
- L16:208 suppresses
- L3:459 suppresses
- L2:238 excites
- L16:48 excites
- L18:127 excites
- L2:598 excites
Feeds later layers
- L27:2661 weakly suppresses
- L27:185 weakly excites
- L27:668 weakly suppresses
- L27:2770 weakly suppresses
- L27:744 weakly excites
- L27:919 weakly suppresses
- L27:2589 weakly suppresses
- L27:765 weakly excites
- L27:1630 weakly suppresses
- L27:1815 weakly suppresses
- L27:1737 weakly excites
- L27:2009 weakly suppresses
- L27:2640 weakly suppresses
- L27:925 weakly excites
- L27:698 weakly suppresses
- L27:847 weakly suppresses
- L27:1384 weakly suppresses
- L27:538 weakly suppresses
- L27:850 weakly suppresses
- L27:453 weakly suppresses
- L27:1525 weakly suppresses
- L27:2897 weakly suppresses
- L27:638 weakly excites
- L27:699 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
- ly
- gli
- ibling
- ughter
- ungle
- imal
- (rawValue
- apor
- теперь
- ive
Suppresses
- 可
- 会让
- 用
- 懂
- 同
- 知
- 出
- 会对
- linear
- 会