reciprocal 'each other' or 'one another'
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. 13 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 "other" or "another" in phrases like "each other" and "one another", which show people or things acting on each other. It is very clear in the examples. When turned up, the model's writing changed topic and did not stay on this idea.
Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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.
-
for a sentence to be grammatically correct, the subject and verb must both be singular or plural. In other words, the subject and verb must agree with one another in their tense. If the subject is in plural form, the verb should also
-
made available to customers using the web. This write-up briefly covers the different types of posting as well as goes over just how these different approaches differ from each various other and just how they can be useful for authors.⏎In a typical publisher-author partnership
-
Pharm.of or pertaining to drugs that interfere with one another chemically or physiologically and therefore cannot be mixed or prescribed together.⏎1
-
nutritional and / or pharmacological.⏎It is very important to note that in addition to external factors, nutritional or pharmacological substances can interact with each other due to its chemical and reactive (acid / base, oxidant / reducer etc
-
task in a foolish manner.⏎3. A father and his sons⏎A father had four sons. He loved them very much. They always used to help one another<|endoftext|>Ramses III, Ramses also spelled Ramesses or Rameses
-
linked may help you understand symptoms and manage them more effectively during this unsettling time.⏎How does it work? The brain and the body are constantly sending messages to each other. These messages tell the brain and body to make changes and adjustments to the way
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 going to write up on the following: I will use the fact that I am going to be thinking about how to explain the fact that I am using a rectangle is like a square is like a rectangle, which is a triangle.
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
a quiver of arrows conspicuously in your hall or library, and you will soon discover that no exquisite painting or bit of statuary will receive from guests more attention than will be accorded to these ancient weapons.⏎No doubt if one could procure
-
對 CNY 感興趣?⏎Early Currency in China⏎With a history of over 300
-
renimine MSÜ Barbaros DEBİM Enstitüsünde devam etmekte aynı zamanda enerji sistem ve modelleme üzerine araştırmalar yapmaktadır.<|endoftext|>- 10 to 15 minutes⏎
-
UNDS ARE THOSE ORGANIC COMPOUNDS WHICH HAVE BENZENE RING (CYCLOHEXA-1,3,5-TRIENE). AROMATIC COMPOUNDS ARE ALWAYS FOLLOWS
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
- L12:2145 excites
- L7:361 excites
- L12:1410 suppresses
- L12:953 weakly suppresses
- L12:2687 weakly suppresses
- L6:628 weakly excites
- L6:2276 weakly suppresses
- L12:2422 weakly excites
- L1:564 weakly excites
- L0:505 weakly excites
- L8:230 weakly excites
- L12:902 weakly suppresses
- L9:1432 weakly excites
- L2:3005 weakly suppresses
- L7:1301 weakly suppresses
- L6:393 weakly suppresses
- L10:2578 weakly excites
- L3:2161 weakly excites
- L1:210 weakly excites
- L9:134 weakly suppresses
- L11:2329 weakly excites
- L5:763 weakly excites
- L6:2883 weakly suppresses
- L11:457 weakly suppresses
Feeds later layers
- L15:853 excites
- L16:2977 excites 'which' after a preposition
- L17:1080 excites
- L16:605 suppresses
- L15:3006 weakly suppresses
- L15:804 weakly suppresses
- L24:545 weakly excites
- L14:539 weakly excites
- L17:850 weakly suppresses
- L15:1614 weakly excites
- L14:1998 weakly excites
- L17:2162 weakly excites
- L17:421 weakly excites
- L16:1365 weakly excites
- L14:2854 weakly excites
- L16:507 weakly excites
- L18:1678 weakly excites
- L15:1744 weakly suppresses
- L17:2878 weakly excites
- L14:1552 weakly suppresses
- L15:2181 weakly excites
- L14:404 weakly excites
- L20:2318 weakly suppresses 'recent' in time phrases
- L24:691 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
- ending
- ine
- Hai
- ег
- esto
- 等级
- 这个
- audiences
- tics
- 端
Suppresses
- 公司将
- Sparse
- _sheet
- _MAY
- icense
- uniqu
- IMO
- Sparse
- baptized
- shape