'Springer' and 'rebirth' style word pieces
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 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 responds to a mixed bunch of words, most often the publisher name Springer, and parts of words like 'rebirth' and 'astronaut'. Turning it up or down did not produce a clear change in what the model wrote.
Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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.
-
, Vol 3118, Springer-Verlag, Berlin⏎Heidelberg New York⏎D., Lenoir
-
, filters, scripts and many basic or more advanced topics. It also includes some tutorials.⏎by Nicholas M. Patrikalakis, Takashi Maekawa - Springer⏎Shape interrogation is the process of
-
from East Side Charter, Springer Middle School, and H.B. duPont Middle School.⏎The project, which
-
had 9 foot ceilings and noticeable window decorations.⏎Gothic architecture is typified by simply the use regarding large, featureless rooftops. The Gothic rebirth style was personified by beautiful monasteries which had huge, domes,
-
03 and repeated it a few times since then.⏎But there were two new dimensions of the recent flight. One, it carried a Chinese woman astronaut into space for the first time. Secondly and far more significantly, it demonstrated China
-
SS: Adaptable⏎The State of the Union Address⏎Introduce your future voters to the annual State of the Union Address. The day after the speech airs on television, use the activity from PBS to discuss the
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 am writing a paper on a research about a study on a paper on a paper on a paper. I need to write a paper on a research paper. I am writing a paper. I need to write a paper. I need to
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
the synergy of every team member.⏎n.大綱, 提要 [ summary or outline of a book, play, etc ]⏎note: 字根
-
Paris to Petersburg in one hour'). The print shows a man seated on a steam-powered rocket-like contraption, which is propelling him through the sky above the rooftops and spires of a town. It is a German version of an English
-
.⏎The root of all energy on Earth is the Sun. The logical source for renewable energy is Solar power. One problem is the size of solar panels - you gotta cover the roof of your house with junk just to have enough umph boil a
-
the impact of the environment.<|endoftext|>5 COMMON CAUSES OF BRAIN FOG⏎You need to stop by the grocery on the way home from work for two gotta-have items.
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
- L3:3037 suppresses
- L21:2252 suppresses
- L19:1016 excites
- L26:2929 weakly excites
- L11:1799 weakly suppresses
- L17:3041 weakly excites
- L1:2094 weakly excites
- L7:2579 weakly excites
- L17:451 weakly suppresses
- L20:2097 weakly suppresses
- L19:1654 weakly excites
- L24:2770 weakly suppresses
- L4:2144 weakly suppresses
- L21:1781 weakly excites
- L18:1955 weakly excites
- L22:2651 weakly excites
- L19:2934 weakly excites
- L17:1863 weakly excites
- L18:1745 weakly excites
- L21:2033 weakly excites
- L13:624 weakly excites
- L25:1138 weakly excites
- L23:2737 weakly excites
- L23:1983 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
- ăm
- 胎
- imposs
- 不可思
- anism
- aż
- mates
- Impossible
- igit
- kil
Suppresses
- ave
- обязательно
- Doc
- 更多
- Prompt
- ernet
- 祚
- AreaView
- olean
- indoors