'of' after a page, article or section number
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. 5 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 word 'of' when it comes right after a number that names a page, article, section or chapter, such as 'page 17 of the textbook' or 'Article 123 of the Constitution'. Turning it up didn't make the text about any particular subject.
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.
-
carriage return (no need to deal with errors, etc.).⏎Reading a character from the keyboard is done via the read_char macro (see page 17 of the textbook for a description). So you'll have to call this macro 5
-
years ago— which the new figures now make out-of-date. (The state’s estimates of sea-level rise and subsidence are listed on page 83 of the Master Plan.)⏎The state plan, while “valuable and thoughtful,” has
-
These holidays are defined in article 18 of the Afghan constitution from 2004. The holidays occur on a fixed
-
? [Hint: remove any blanks from each column, remembering to sort the columns together.] Use an F test (on pages A-2 and A-3 of your Class Notes book). (You are advised NOT to use the Two-sample F
-
.⏎We need to grumble and complain about drivers who fail to use good winter tires and who overdrive the conditions.<|endoftext|>Viewing page 28 of 101⏎It looks like you're using a mobile device. We
-
country and for restoring Mexico to a Republic.⏎May 1st: Labor Day, commemorates the union movement of the Mexican worker. Article 123 of the Constitution of 1917 gave workers the right to organize labor unions
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 need to get this problem to be. So,, I need to get this problem to be,, I need to get,,, and the, and, and others, and the, and, and others, and the,
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
played in time (that is what rhythm is!!).⏎Many people count sixteenth notes this way: ‘one ee and a, two ee and a, three ee
-
Start with the sentence ‘Like the Corelli, the harmony in Haydn’s Joke quartet is mainly diatonic and functional but with some occasional dissonance’.⏎No homework this week⏎For Friday 12th – Finish
-
last bar, each note is played on each beat.⏎There are a few different ways of counting eighth and sixteenth notes.⏎For eighths notes you can count ‘1 and 2 and
-
hour.⏎Pythagoras' concepts included the mathematical order of the cosmos, and he may have been led to this assessment from the mathematical order of music (consonants of octaves, fifths, and fourths being produced by simple ratios
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
- L3:2441 excites
- L4:2255 weakly suppresses
- L0:1450 weakly excites
- L3:2928 weakly excites
- L4:1140 weakly excites
- L2:521 weakly suppresses
- L4:906 weakly excites
- L3:797 weakly excites
- L1:2061 weakly suppresses
- L3:2655 weakly excites
- L1:1226 weakly excites
- L0:1545 weakly excites
- L3:616 weakly suppresses
- L1:2084 weakly suppresses
- L3:856 weakly suppresses
- L2:112 weakly excites
- L3:286 weakly excites
- L4:2697 weakly suppresses
- L0:392 weakly suppresses
- L1:3071 weakly suppresses
- L3:72 weakly suppresses
- L2:59 weakly excites
- L3:1135 weakly excites
- L2:996 weakly excites
Feeds later layers
- L9:602 excites
- L12:752 excites
- L16:3051 weakly suppresses
- L9:673 weakly excites
- L7:1941 weakly excites
- L20:2672 weakly suppresses
- L9:942 weakly excites
- L6:336 weakly excites
- L6:998 weakly suppresses
- L6:345 weakly excites
- L10:1626 weakly excites
- L12:203 weakly excites
- L14:629 weakly suppresses
- L21:893 weakly excites
- L21:1702 weakly suppresses
- L20:1236 weakly excites
- L13:1925 weakly suppresses
- L18:925 weakly suppresses
- L17:2109 weakly suppresses
- L12:1594 weakly suppresses
- L7:2784 weakly suppresses unclear: 'Match' in exercise instructions
- L6:2217 weakly excites
- L11:1147 weakly excites
- L9:926 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
- kip
- kop
- ultipartFile
- overall
- 的整体
- destruction
- highest
- lips
- Lips
- JSImport
Suppresses
- 品牌
- ensible
- 孩子
- elter
- 名字
- 填补
- iej
- rieved
- fname
- ahun