'of' after abstract nouns like 'sense' or 'view'
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. 2 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 responds to the word 'of' that follows nouns such as 'sense', 'view', 'need' or 'idea', as in 'sense of injustice'. Its negative side picks up words like 'inception' or 'absence' that follow 'their' or 'since'. Turning it up did not give a clear theme in the model's writing.
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.
-
glimpse into the academic world and its view of religion in general and Hinduism in particular. The hundred-plus pages of articles are
-
form of English was essentially a request from the customers (i.e. the Airlines), of which 80% are not native English speakers, and their need of doing correct maintenance to guarantee the aircraft availability.⏎Complex technical instructions can be misunderstood and
-
stand to see animals killed just for fun. Her sense of injustice makes her angry and through her anger she develops a special power - not diss
-
?⏎- Do the characters/narrator reflect upon whether they exist and how they know what they know?⏎- To what extent do the following factors impact their sense of self: emotions, past
-
small as asteroids go. Because of that (and its current distance of 450 million kilometers) it's a bit faint, shining right now
-
or with a jacket when going out, the shirts relaxed basket weave also allowed them to be worn pretty much year round. Today, practically every company produces their version of the OCBD, and they’re somewhat of a collectors item, not because they
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 have been thinking about lately: I have been thinking about lately: I have been thinking about lately: I have been thinking about lately: I have been thinking about lately: I have been thinking about lately: I have been thinking about lately
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
Europe where it formed an intricate part of military bands. Pipe and drum bands were quite popular and predate military bands with a wide variety of instruments. Since their inception, snare drums can be heard in virtually all types of European and North American
-
ment. This has now been banned, captive breeding sites for each species has been established, but the problems involved in saving these birds are immense.⏎In their absence, populations of feral dogs and rats have taken over their scavenging role
-
is subject to mechanical forces from intramolecular tension? An important component of the proposed project is technology development, which has been a hallmark of this grant since its inception in the 1980s, when we established the first quantitative in
-
take up torches and pitchforks against him on the silver screen. Never once during his rampage, however, did the monster question the honesty of his ultimate creator,
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
- L0:326 weakly excites
- L1:993 weakly suppresses
- L1:433 weakly suppresses
- L1:2902 weakly suppresses
- L1:2362 weakly excites
- L1:2979 weakly suppresses
- L1:2801 weakly excites
- L0:1523 weakly excites
- L1:454 weakly excites
- L1:370 weakly excites
- L0:414 weakly suppresses blank after a dash in number ranges
- L0:121 weakly excites
- L1:1993 weakly excites
- L0:620 weakly suppresses
- L1:1340 weakly suppresses
- L1:2925 weakly excites
- L1:383 weakly excites
- L0:1745 weakly excites quantity word after large, high or low
- L1:426 weakly excites
- L0:1750 weakly excites
- L0:1878 weakly excites
- L1:1419 weakly excites
- L0:670 weakly suppresses
- L1:752 weakly excites
Feeds later layers
- L3:4 excites
- L7:1821 suppresses
- L12:575 excites
- L7:1624 suppresses
- L9:69 excites
- L4:81 suppresses
- L14:282 excites
- L4:510 excites
- L3:1015 suppresses
- L9:537 excites
- L4:72 suppresses
- L22:1051 suppresses
- L3:58 weakly excites
- L3:402 weakly suppresses
- L9:1182 weakly suppresses
- L5:453 weakly suppresses
- L9:1412 weakly suppresses the word "be" after modals and "to"
- L10:235 weakly excites
- L6:2221 weakly suppresses
- L6:285 weakly suppresses
- L5:305 weakly excites
- L3:471 weakly suppresses
- L5:804 weakly suppresses
- L6:101 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
- -type
- type
- -Type
- -lo
- ชน
- ươi
- -rel
- lo
- tipo
- -text
Suppresses
- ');</
- acie
- _history
- atIndex
- 灯火
- anale
- 历史上
- zza
- 髯
- 琬