dash starting a bullet item
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. 11 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 weakly lights up on the dash that starts a new bullet point in a list of tips or notes, and on phrases like 'For instance'. The strongest negative cases are at end-of-document markers, and the steered text did not show a clear theme.
Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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.
-
Thermostats regulate the system’s temperature, while filters clean the circulated air. Understanding these parts is crucial for effective maintenance and troubleshooting. For instance, if a room isn’t reaching the desired temperature, the problem could lie with
-
ant that contains at least 70% alcohol.⏎- Other recommendations⏎- Avoid keeping your smartphone and handkerchief in the same pocket
-
2-82571; Revised: 22-Nov-2022, Manuscript No. AAAJMR-22-82571(R); Published: 29-Nov-2022
-
faucets to avoid wasting water.⏎- Consider upgrading to a more efficient water heater.⏎Cook with energy savings in mind.
-
substitution⏎- Write the meaning of Centrally Planned Economy⏎- What is demand function?⏎
-
not regular.⏎- You need a diphtheria booster shot.⏎- You have questions or concerns
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 dog." The first thing that comes to mind is that I'm trying to get a dog to understand at around 10:00 a.m. when the sun is already up. I've been writing that as
Most negative on
Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.
-
support to families living with type 1 diabetes.⏎© Children with Diabetes, Inc. 1995-2017. Comments and Feedback.<|endoftext|>Scoliosis vs. Ky
-
guide only. We cannot offer interpretations about citations as this is an automated procedure. Please refer to the style manuals in the Citation Source Information area for clarifications.<|endoftext|>The ceramic firing technique termed "raku" was developed in Japan during the
-
Back to MedTerms online medical dictionary A-Z List⏎Need help identifying pills and medications?⏎Get the latest health and medical information delivered direct to your inbox FREE!<|endoftext|>November 16, 2006⏎- by Robert E.
-
/the-omnivores-delusion-against-the-agri-intellectuals/⏎Leave a ReplyWant to join the discussion?⏎Feel free to contribute!<|endoftext|>Tree Roots Can Wreak Havoc on Pipes — How to Best Prevent Pipe Damage
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
- L10:710 strongly excites
- L10:1298 strongly excites
- L9:125 excites
- L10:1935 suppresses
- L8:313 suppresses
- L8:624 excites
- L10:2423 suppresses
- L8:269 excites
- L10:2061 suppresses
- L10:2475 excites
- L9:1033 suppresses unclear: end-of-text markers and word fragments
- L9:2166 excites
- L8:1090 excites
- L10:1714 excites
- L7:2961 suppresses
- L10:851 weakly excites
- L8:2725 weakly excites line-break tokens inside wrapped text
- L7:542 weakly suppresses
- L8:1207 weakly excites
- L9:1045 weakly excites
- L3:829 weakly suppresses
- L7:1707 weakly excites
- L3:2329 weakly suppresses
- L9:1633 weakly suppresses
Feeds later layers
- L13:892 excites end of text token, weak
- L14:3021 excites
- L12:2927 excites
- L14:1182 excites
- L15:2008 suppresses
- L17:2032 excites
- L26:2128 excites decimal points inside numbers
- L12:978 suppresses
- L13:1152 excites
- L12:1309 excites
- L12:1313 suppresses
- L16:2575 suppresses
- L12:864 weakly excites
- L13:1038 weakly excites
- L14:999 weakly suppresses
- L14:1504 weakly excites
- L12:543 weakly suppresses
- L22:717 weakly excites
- L13:2673 weakly suppresses
- L12:2368 weakly excites
- L16:2664 weakly excites
- L19:2057 weakly excites
- L15:1748 weakly suppresses
- L15:506 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
- itest
- '{@
- 狁
- Qgs
- inue
- bond
- QAction
- glEnd
- UCCESS
- supper
Suppresses
- 翻开
- рож
- definitions
- DESCRIPTION
- изд
- Overview
- فة
- definition
- 两种
- 开