"to" after adjectives like obvious or meaning
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. 6 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 "to" when it follows words like "obvious", "meaning" or "different", as in "obvious to me" or "what freedom means to me". Turning it up made the text wander into unrelated topics.
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.
-
Susan L. Holdt, Bertrand Jacquemin and Céline Rebours<|endoftext|>Performing well in college means everything to college students.⏎So much so, that as many as 87% of
-
was no color bar at Cooper Union. Cooper demanded only a willingness to learn and a commitment to excellence, and in this he manifestly succeeded.⏎It was obvious to Cooper that not everyone would be able to enroll
-
unfortunate. Software patents impede innovation, and don’t benefit the public. If you find prior art that shows any of this method is not new or is obvious to a person trained in the art, then let the world know.<|endoftext|>Should you
-
this essay would be for her. She simply wrote the following from her heart:⏎Freedom is a word that has different meanings to everyone. So what does freedom mean to me? Well, it is the describing word for our country. It’s the
-
the movie my notions were challenged when I noticed that this video was going to be silent. It only happened to me for a couple seconds and then it became obvious to me that there was no other way to make a video about this matter than sound
-
es⏎- Back hole, arse, butt, bussy⏎- Breasts, chest, breast tissue, boobs, tits⏎Sex can mean different things to different people. Sex can include vaginal, anal, oral, hand jobs or fing
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 a question about the number of times the number of times the number of times the number of times the number of times the number of times the number of times the number of times the number of times the number of times the number
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
DataTableCollection. Dim tablesCollection As DataTableCollection = thisDataSet.Tables ' Get a specific table by name. Dim table As DataTable = tablesCollection("Suppliers") Console.WriteLine(table.TableName) End Sub⏎Available since 1.1<|endoftext|>A musical
-
sectable) and who have symptoms like weight loss, nausea, buildup of fluid in the abdomen (ascites) and a poor⏎Margins in surgery @(Model.HeadingTag)>⏎Margin is the area of healthy tissue surrounding a tumour that is
-
teachers must demonstrate careful planning, challenging goals, creative use of resources, varied and effective teaching strategies and awareness of their own strengths and weaknesses.⏎The standards set the bar high because the<|endoftext|>At the higher data link layer of communication, asynchrony
-
ise symptoms in a Lancet paper. At a much less sophisticated level, PutnamPig’s “Swine Flu Song” on Youtube http://www.youtube.com/watch?v=tbt_PuVAVTU defends pigs against unfair stigmatisation
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
- L4:1765 excites
- L4:1183 weakly suppresses
- L3:1229 weakly excites
- L4:949 weakly excites
- L3:213 weakly suppresses
- L4:1896 weakly suppresses
- L1:1766 weakly suppresses
- L4:792 weakly excites
- L0:2594 weakly suppresses
- L1:1674 weakly excites
- L1:257 weakly suppresses
- L0:1672 weakly suppresses
- L5:1524 weakly suppresses
- L5:1376 weakly suppresses
- L4:939 weakly excites
- L2:3033 weakly excites
- L5:2251 weakly suppresses
- L4:271 weakly excites
- L3:1865 weakly suppresses
- L3:1920 weakly excites
- L1:1367 weakly suppresses
- L0:2432 weakly suppresses
- L5:1409 weakly suppresses
- L2:3039 weakly suppresses
Feeds later layers
- L15:451 suppresses
- L11:1047 suppresses pieces of software and library names
- L7:304 excites
- L8:509 suppresses
- L18:274 suppresses
- L10:457 excites
- L21:2438 excites
- L9:226 suppresses 'to' starting an infinitive
- L9:1627 excites
- L26:970 weakly excites
- L19:913 weakly suppresses
- L19:745 weakly suppresses
- L14:1248 weakly suppresses
- L13:1401 weakly suppresses
- L25:343 weakly excites
- L19:1596 weakly suppresses
- L7:2766 weakly excites
- L21:1476 weakly excites
- L10:678 weakly suppresses
- L17:742 weakly suppresses
- L17:1674 weakly suppresses
- L9:654 weakly excites
- L16:2366 weakly suppresses
- L7:26 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
- essor
- essional
- orman
- tant
- 各行
- ]>⏎
- 你自己
- 毕竟是
- 蒟
- 大型
Suppresses
- 善
- .d
- .de
- ::
- -icons
- .uni
- /icons
- 마
- 원
- .get