opening parenthesis before an acronym
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. 10 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 lights up on an opening bracket that comes right after a full name, just before its short form is written in brackets, such as "Graphical User Interface (GUI)". Turning it up did not produce a clear visible effect on the 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.
-
Dendrophthoe|⏎|Leaves opposite or clustered; inflorescence a simple<|endoftext|>Computer scientists concerned with the systems security of Direct-Recording Electronic (DRE) voting machines are by now familiar with the Diebold debacle of
-
credits sourced from sustainable practices in climate-affected households in Mongolia. The project's digital MRV (measurement<|endoftext|>In this blog post we will use a Python script to work out how
-
access to information and communication technologies (ICTs) can mitigate many issues in SIDS and help to diversify their
-
certainly supports our point of view that in the grief of a departure, wailing is a natural habit. Wailing or weeping is a natural instinct hence Allah (swt) has
-
. In order to create a balance, both husband and wife enjoy their own well-defined rights in their own capacities, as quoted in the Holy Quran:⏎“And they (women) have rights (over their husbands) similar (to those of their husbands
-
another mother⏎The concept of digital twins, particularly when combined with MSPC, is relatively new, yet a well-constructed system based on Process Analytical Technology (PAT) will leverage both of these concepts, and as a result maximize your process
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 most important thing to know the other two things, and the third thing, and the fourth, and the fifth, and the sixth, and the seventh, and the eighth, and the ninth, and the tenth, and the ele
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
||CRC Press|⏎|Date published: || 2015|⏎|ISBN: ||9781498738552 (DRM-PDF)|⏎|Read Aloud: ||not allowed|<|endoftext|>
-
In 1949, the right to asylum was laid down in the Constitution. Until 1993, Article 16 of the Constitution (Grundgesetz)
-
to the FCC at 888-225-5322, at 888-835-5322 (TTY) or at email@example.com.<|endoftext|>Table of Contents :⏎Top Suggestions Senior
-
first naval battle in which victory was decided by aircraft alone, Dauntless dive-bombers fought well alongside other US aircraft. Each side lost one carrier (the Japanese carrier Shoho being sunk by Dauntless and
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
- L9:1414 suppresses opening parenthesis after copyright or year
- L9:1065 suppresses
- L8:618 excites
- L8:185 excites
- L8:131 excites
- L9:2964 excites
- L9:1370 suppresses
- L5:1847 weakly suppresses
- L8:666 weakly excites
- L4:2802 weakly suppresses
- L5:114 weakly suppresses
- L3:173 weakly suppresses
- L7:685 weakly excites
- L6:804 weakly excites opening bracket before an acronym
- L8:1097 weakly suppresses
- L7:1302 weakly suppresses
- L7:724 weakly excites
- L8:70 weakly suppresses 'and' between stacked superlatives
- L4:576 weakly suppresses
- L8:577 weakly excites
- L9:2259 weakly suppresses
- L5:1061 weakly suppresses
- L6:1392 weakly excites
- L7:135 weakly excites 'of' after 'issue' in citations
Feeds later layers
- L18:2544 excites
- L16:976 excites
- L12:3064 excites
- L17:920 excites
- L16:2969 excites
- L15:2579 excites
- L15:1576 weakly excites
- L15:1570 weakly excites
- L16:1573 weakly excites
- L12:3070 weakly suppresses
- L15:582 weakly excites
- L12:1197 weakly suppresses
- L12:2132 weakly excites
- L17:444 weakly excites
- L12:432 weakly excites
- L13:820 weakly suppresses comma after an introductory phrase
- L15:2884 weakly excites
- L12:2208 weakly suppresses
- L17:795 weakly suppresses
- L13:114 weakly excites
- L14:1181 weakly excites
- L19:1858 weakly excites
- L13:631 weakly suppresses
- L15:2437 weakly excites opening parenthesis after a noun
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
- zew
- zyst
- statt
- there
- sert
- cor
- ort
- =tk
- there
- ╔
Suppresses
- 部主任
- ルド
- idential
- ials
- *</
- iline
- 过的
- eware
- 重中之重
- 合理的