Neuron Explorer

Qwen3-0.6B

dash starting a bullet item

A neuron in layer 11 of Qwen3-0.6B. Its strongest snippets are very consistent: they mostly peak on the same token. It is 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 activation was highest. Amber marks positive activation; the outlined token is the peak. Hover a token for its strength.

  1. #1 peak

    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

  2. #2 peak

    ant that contains at least 70% alcohol.⏎- Other recommendations⏎- Avoid keeping your smartphone and handkerchief in the same pocket

  3. #3 peak

    2-82571; Revised: 22-Nov-2022, Manuscript No. AAAJMR-22-82571(R); Published: 29-Nov-2022

  4. #4 peak

    faucets to avoid wasting water.⏎- Consider upgrading to a more efficient water heater.⏎Cook with energy savings in mind.

  5. #5 peak

    substitution⏎- Write the meaning of Centrally Planned Economy⏎- What is demand function?⏎

  6. #6 peak

    not regular.⏎- You need a diphtheria booster shot.⏎- You have questions or concerns

Steering

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.

turned down a lot
Prompt

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.

  1. #1 peak

    support to families living with type 1 diabetes.⏎© Children with Diabetes, Inc. 1995-2017. Comments and Feedback.<|endoftext|>Scoliosis vs. Ky

  2. #2 peak

    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

  3. #3 peak

    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.

  4. #4 peak

    /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 map

Which neurons in earlier layers feed this one, and which later ones it feeds. Amber links excite, blue links suppress; longer bars are stronger.

Fed by earlier layers

  1. L10:710 strongly excites
  2. L10:1298 strongly excites
  3. L9:125 excites
  4. L10:1935 suppresses
  5. L8:313 suppresses
  6. L8:624 excites
  7. L10:2423 suppresses
  8. L8:269 excites
  9. L10:2061 suppresses
  10. L10:2475 excites
  11. L9:1033 suppresses unclear: end-of-text markers and word fragments
  12. L9:2166 excites
  13. L8:1090 excites
  14. L10:1714 excites
  15. L7:2961 suppresses
  16. L10:851 weakly excites
  17. L8:2725 weakly excites line-break tokens inside wrapped text
  18. L7:542 weakly suppresses
  19. L8:1207 weakly excites
  20. L9:1045 weakly excites
  21. L3:829 weakly suppresses
  22. L7:1707 weakly excites
  23. L3:2329 weakly suppresses
  24. L9:1633 weakly suppresses

Feeds later layers

  1. L13:892 excites end of text token, weak
  2. L14:3021 excites
  3. L12:2927 excites
  4. L14:1182 excites
  5. L15:2008 suppresses
  6. L17:2032 excites
  7. L26:2128 excites decimal points inside numbers
  8. L12:978 suppresses
  9. L13:1152 excites
  10. L12:1309 excites
  11. L12:1313 suppresses
  12. L16:2575 suppresses
  13. L12:864 weakly excites
  14. L13:1038 weakly excites
  15. L14:999 weakly suppresses
  16. L14:1504 weakly excites
  17. L12:543 weakly suppresses
  18. L22:717 weakly excites
  19. L13:2673 weakly suppresses
  20. L12:2368 weakly excites
  21. L16:2664 weakly excites
  22. L19:2057 weakly excites
  23. L15:1748 weakly suppresses
  24. L15:506 weakly excites

Activation histogram

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.

2 values between -2.04 and -1.952 values between -1.24 and -1.151 values between -1.15 and -1.062 values between -1.06 and -0.981 values between -0.98 and -0.894 values between -0.89 and -0.804 values between -0.80 and -0.7111 values between -0.71 and -0.627 values between -0.62 and -0.5321 values between -0.53 and -0.4434 values between -0.44 and -0.3553 values between -0.35 and -0.27127 values between -0.27 and -0.18332 values between -0.18 and -0.091252 values between -0.09 and 0.001338 values between 0.00 and 0.09419 values between 0.09 and 0.18187 values between 0.18 and 0.2769 values between 0.27 and 0.3622 values between 0.36 and 0.458 values between 0.45 and 0.532 values between 0.53 and 0.624 values between 0.62 and 0.711 values between 0.71 and 0.802 values between 0.80 and 0.891 values between 0.89 and 0.981 values between 1.33 and 1.421 values between 1.42 and 1.51offnegativepositiveoften

Logit lens

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 layers instead.

Promotes

  1. itest
  2. '{@
  3. 狁
  4. Qgs
  5. inue
  6. bond
  7. QAction
  8. glEnd
  9. UCCESS
  10. supper

Suppresses

  1. 翻开
  2. рож
  3. definitions
  4. DESCRIPTION
  5. изд
  6. Overview
  7. فة
  8. definition
  9. 两种
  10. 开