comma before a participle phrase
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. 20 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 a comma that comes after a clause like "and" or "then", just before the sentence continues with a phrase such as "finding his enemy dead". Turning it up made the model stutter on the word "using", while turning it down gave ordinary text.
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.
-
by Julius Caesar in Greece, fled to Egypt seeking solace but was immediately murdered by agents of Ptolemy XIII. Caesar arrived in Alexandria soon after and, finding his enemy dead, decided to restore order in Egypt.⏎During the preceding century
-
amusement.⏎There was a half-decayed log lying by the roadside, and Grant sat down on it, calmly pulled a cigar from his cigar case, and, requesting one of his staff who stood nearby to furnish him a sheet of paper,
-
Hebrides that Columba turned his face when he accepted the terrible penance of Molaise. He bade farewell to his relatives, and, with a few monks who insisted on accompany him whithersoever he might
-
put his scalp on display for money.⏎In 1900, Thompson sent the scalp back to the Nebraska doctor who tried to reaffix it, who, in turn, donated it to the public library. The scalp remained on display there
-
posited answers to this question, whether directly - in the case of Plato and Plotinius - or through story - in the case of Homer.⏎Plotinus, having been a proponent of Platonism, naturally concurs with Plato,
-
built except the roof," Thomas replied. So Gundaphorus sent him gold and silver to roof the palace, and Thomas, thanking<|endoftext|>The Internet and copyright⏎Choose a topic:⏎Internet technology is developing faster
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: the design of a software system which is not only created to create but also to create create... create and create create... create create. It is so created, so create and create. It is also created to create create... create create.
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
,0<|endoftext|>Guidelines for Restoring Cultural Landscapes⏎Identifying, retaining and preserving
-
1,700.<|endoftext|>Risk assessment for construction⏎The Construction (Design and Management) Regulations (CDM
-
A white turkey is a domesticated turkey that has been bred to have white feathers. White turkeys are not albino; they have black eyes and dark skin. The Leycester sub-group of the Bronze turkey is the only true genetic white
-
others only shame and hardship.⏎This I can say for myself:⏎that for awhile I was the Heodeninga<|endoftext|>This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.⏎As
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
- L19:94 suppresses
- L19:2683 suppresses
- L19:2709 excites
- L18:2434 excites
- L17:3068 excites comma after a sentence subject
- L19:788 weakly excites
- L15:835 weakly suppresses
- L16:1808 weakly excites
- L11:2200 weakly excites
- L2:2775 weakly excites
- L19:2707 weakly excites
- L17:2041 weakly suppresses
- L0:509 weakly suppresses
- L4:1079 weakly suppresses
- L19:1819 weakly excites
- L19:119 weakly suppresses
- L17:1209 weakly suppresses
- L18:1926 weakly suppresses
- L12:1445 weakly suppresses
- L11:1045 weakly suppresses
- L9:1137 weakly excites
- L3:2958 weakly excites
- L17:1839 weakly suppresses
- L3:1908 weakly suppresses
Feeds later layers
- L24:140 strongly excites
- L22:1425 strongly excites
- L25:376 strongly excites
- L26:252 strongly excites
- L26:686 strongly excites
- L27:410 strongly excites
- L27:1237 strongly excites
- L22:609 excites
- L27:1890 excites
- L25:1596 excites
- L27:303 excites
- L27:1148 excites
- L25:218 excites
- L26:708 excites
- L23:424 excites
- L25:620 excites
- L27:677 excites
- L26:1304 excites
- L27:1545 excites
- L21:1630 excites
- L21:615 excites
- L27:1852 excites
- L26:733 excites
- L26:852 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
- using
- using
- -using
- relying
- applying
- considering
- _using
- taking
- recognizing
- preferring
Suppresses
- nữa
- ocumented
- 其
- duce
- 那就是
- create
- Describe
- cả
- CloseOperation
- Creates