unclear: word starts like Ag and Re
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 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 the first piece of certain names and words, such as 'Ag' in Agatha or Agudat, and 'eg' in Aedes aegypti. Its strongest negative cases are the start of words like 'Reproduced'. The model's output under steering was broken, so the role is not clear.
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.
-
disagreements over religious observances and liturgy, a group of more traditional members left to establish a separate Orthodox service. In time, this group evolved into the congregation Agudat Achim.⏎During this same period, B’nai Israel was finding
-
as 850 meters or half a mile or so (see P. Reiter et al., "Short Report: Dispersal of Aedes Aegypti in an Urban Area After Blood Feeding as Demonstrated by Rub
-
smell what The Rock is cooking?" He used illeism deliberately to exaggerate his self-importance.⏎Think also of the character Hercule Poirot in Agatha Christie’s mysteries. Christie often portrayed the detective as referring to himself in the
-
mother of Geb and Nut.⏎Tefnut and her husband Shu were the children of Atum who created them by masturbating. When Atum became associated with Re, Re became Tefnut's father. According to one myth, Tef
-
let you off the hook, even if you got 98 out of 100 on an exam.⏎Back to top⏎The Role of Indirect Aggression⏎What goes on outside the home is also a big factor in the development
-
speed wind streams in solar and stellar coronal holes, and plasma heating in closed magnetic structures. Dr. Davila<|endoftext|>Life Cycle and Information on Aedes Aegypti Mosquito
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:dd cde dde dde cde cde cde cde cde cde cde cde cde dde cde cde
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
.<|endoftext|>MLA style for documentation is widely used in the humanities, especially in writing on language and literature.⏎One standard citation format applies to every source type⏎Reproduced from https://style.mla.org/whats-new/⏎Author.
-
29 Gastrooesophageal reflux disease⏎Musculoskeletal system 32 Osteoporosis⏎Endocrine system 35 Diabetes mellitus⏎Reproductive system 37 Erectile dysfunction⏎
-
og, acid rain, carbon monoxide, fossil fuel exhausts, and tropospheric ozone have in common? They are all examples of air pollution.<|endoftext|>Revealing the art and architecture of Vijayanagara Empire, Vittala Temple
-
improved infrastructure and higher wages isn’t cheap for smaller companies like his. The result, he says, is “a glut of skilled workers on the streets now.”⏎Reactions by western garment industry leaders
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
- L23:1695 suppresses
- L21:162 excites
- L18:335 weakly suppresses
- L24:1232 weakly excites
- L22:1392 weakly excites
- L6:2195 weakly excites
- L7:2095 weakly suppresses
- L6:2341 weakly excites 'mean' after 'doesn't' in 'just because' sentences
- L1:2915 weakly excites
- L2:1745 weakly excites
- L23:1621 weakly excites
- L22:2003 weakly suppresses
- L17:389 weakly excites
- L16:83 weakly excites
- L14:1929 weakly excites
- L19:377 weakly suppresses
- L5:236 weakly excites
- L19:2996 weakly excites
- L6:413 weakly excites
- L17:1627 weakly excites
- L4:1197 weakly suppresses
- L10:269 weakly suppresses joining word in a list
- L10:2909 weakly excites
- L15:2634 weakly suppresses
Feeds later layers
- L27:552 strongly suppresses
- L27:310 suppresses
- L27:845 suppresses
- L27:1680 weakly suppresses
- L27:343 weakly excites
- L27:1683 weakly excites
- L27:305 weakly suppresses
- L27:1905 weakly excites
- L27:2496 weakly suppresses
- L27:2948 weakly suppresses
- L27:1579 weakly excites
- L27:1284 weakly excites
- L27:2686 weakly excites
- L27:2570 weakly suppresses 'con' starting a word piece
- L27:2915 weakly excites
- L27:557 weakly suppresses
- L27:2504 weakly excites
- L27:2585 weakly excites
- L27:1713 weakly suppresses
- L27:1522 weakly excites
- L27:833 weakly excites
- L27:2497 weakly excites
- L27:1346 weakly excites
- L27:837 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
- 版权归原
- ưới
- acci
- ights
- 亭
- итель
- eger
- Nest
- NEL
- 簽
Suppresses
- d
- d
- *d
- .d
- >d
- "d
- [d
- д
- <d
- ,d