opening parenthesis after a term
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
Fires on the opening bracket or colon that comes right after a term and is about to explain it, such as a word followed by its meaning in brackets. Turning it up gave repetitive or rambling text that does not connect clearly to this pattern.
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.
-
among their many strange⏎body-shifting traits. Leaders of Mexican fisheries and scientists⏎are banding together to figure out whether the diablos⏎rojos (red devils
-
decades to the Sun Belt, where homes are cheap and jobs are plentiful but climate risks mount by the day? Or, following the logic of Biden’s own order, should funds be directed to more resilient regions where people might move someday? If it
-
00⏎American History / Women's Studies/Gender Studies⏎View main book page⏎Table of Contents⏎Chapter 1. Sleep Is for Sissies: Elite Males as Paragons of Wakefulness⏎Chapter 2.
-
the pointers is larger, this value may be positive or negative.<|endoftext|>Think of your body like a car. Where the muscles are the shock systems and the skeleton (joints/bones) is the frame of the car. The job of the
-
is hoped can ultimately influence public health policy. Tibetan mothers increased the practice of breastfeeding during the project. However, traditional customs of supplementing breast milk with tsampa (barley porridge) are deeply engrained in the mothers and family advisors,
-
sessions (majalis) in meeting places known as zawiyas, khanqahs or tekke. They strive for ihsan (perfection of worship), as detailed in a hadith: “Ihsan
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 Aha! So I'm not sure, so I have to. Let me. So I have to. Let me. I'm. So. I'm. So. So. Let me. So. Let. So. So
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
shown to be generally safe, citing statistics on how many Americans have died after the seasonal influenza vaccine in recent years.⏎Each year, more than 165 million Americans get the flu shot, and according to the U.S. vaccine adverse event
-
his scholarship was withdrawn for lack of funds. “I thought the world was coming to an end,” the 97-year-old says. Mary Downey ’41 agrees: “People lost everything back then, their jobs and their savings,
-
ysis of human red blood cells would occur if the cell were in .. solution A) isotonic B) hypertonic C) hypotonic D) none of these I think it's B. I'm not too sure though.⏎November 2
-
is not improved by medication or therapy.⏎Share this article⏎Depression is one of the most common forms of mental illness, affecting more than 350 million people worldwide, according to a Reuters video. Bob Holmes is one of them.
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
- L12:2240 suppresses
- L11:2719 suppresses
- L7:2514 suppresses
- L6:1389 excites
- L10:271 excites
- L11:547 excites
- L11:2066 suppresses
- L12:2947 excites
- L9:1065 suppresses
- L8:933 suppresses words introducing an example or aside
- L9:1414 suppresses opening parenthesis after copyright or year
- L8:747 weakly excites
- L0:866 weakly suppresses
- L9:1370 weakly suppresses
- L7:2085 weakly suppresses
- L7:3058 weakly excites
- L6:990 weakly excites
- L6:428 weakly excites
- L7:508 weakly excites
- L12:2171 weakly suppresses
- L11:408 weakly excites
- L12:2293 weakly excites
- L11:2166 weakly excites
- L11:498 weakly excites
Feeds later layers
- L14:2755 suppresses
- L16:2663 suppresses
- L14:2820 suppresses "and" or "or" linking two counted items
- L15:2407 suppresses
- L16:3011 suppresses
- L17:1869 suppresses
- L15:2997 suppresses
- L16:2917 suppresses
- L14:2116 weakly suppresses
- L16:123 weakly excites
- L16:978 weakly suppresses
- L14:2006 weakly suppresses
- L15:2198 weakly suppresses
- L19:1125 weakly excites
- L16:2105 weakly suppresses
- L18:119 weakly excites
- L15:2720 weakly suppresses
- L22:1716 weakly suppresses
- L18:576 weakly suppresses
- L16:268 weakly excites
- L15:2499 weakly suppresses
- L15:1236 weakly suppresses
- L15:1547 weakly suppresses "also" before a verb
- L14:1933 weakly suppresses
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
- agraph
- �
- 蛉
- -kind
- :⏎
- predicted
- ıyla
- bare
- extensively
- Scrollbar
Suppresses
- ains
- aining
- 制
- teness
- aways
- apsed
- AIN
- ระว
- /apt
- =j