Neuron Explorer

Qwen3-0.6B

inside company and brand names

A neuron in layer 12 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on the later part of a company, brand, or organisation name, such as 'Elsevier', 'Mendel Law', or 'Kopila Nepal'. It is most negative inside usernames and web addresses. Turning it up did not produce any clear change in topic that matches 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 activation was highest. Amber marks positive activation; the outlined token is the peak. Hover a token for its strength.

  1. #1 peak

    ability to think and reason. PD can affect the gastrointestinal system and autonomic nervous system, leading to changes in gut health, sleep and mood (Parkinson’s Australia Inc., 2014). Recognising these signs and symptoms can aid

  2. #2 peak

    cruel standard for down production, many brands have adopted the Responsible Down Standard and other certifications, including the Canadian Down & Feather Company, Coyuchi, Downlight Bedding, North Face, H&M, Patagonia and IKEA. A number

  3. #3 peak

    monitor the trends in NCD prevalence, their risk factors, and treatments.⏎Copyright © 2020 The Royal Society for Public Health. Published by Elsevier Ltd. All rights reserved.<|endoftext|>We all grew up hearing the phrase, “

  4. #4 peak

    time. Just add a new Alka-Seltzer tablet and watch as the colored bubbles rise again!⏎2. Lemon Volcano⏎Image Credit: sixthbloom.com⏎Lemon volcanoes don’t create a gigantic eruption, but rather small

  5. #5 peak

    Established in 2001 Working with Children and Women in Nepal⏎2 Kopila Nepal Established in 2001 Working with Children and Women in Western Region of

  6. #6 peak

    you may want to know more about the lifespan of the trust. The Houston living trust lawyers at Mendel Law Firm, L.P. explain when and how a living trust terminates.⏎What Is

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 been thinking about for a long time, but I have been thinking about all the time I have been thinking about all the time I have been thinking about all the time except I have been thinking about all the time except I have been

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

    study for the Moxy people. I did a post and review on that article. https://t.co/D9NIorgXuc— Bruce Rogers (@bjrmd) August 30, 2021⏎PACING:

  2. #2 peak

    219 Matrices I wrote (improved and extended version of JonA1961’s MAX7219array https://github.com/JonA1961/MAX7219array )<|endoftext|>Rein

  3. #3 peak

    The Framing Camera project is funded by the Max Planck Society, DLR, and NASA/JPL.⏎More information about Dawn is online at http://dawn.jpl.nasa.gov.⏎Image credit: NASA/JPL-Caltech/UCLA/M

  4. #4 peak

    list. Other tools on the Voyant site are Word links (creates a mind map of words) and a scatter plot (for word frequencies). http://voyant-tools.org/⏎Wordle: Wordle is a free web<|endoftext|>His wife once

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. L10:2956 suppresses
  2. L10:893 suppresses
  3. L9:736 suppresses
  4. L10:2102 excites
  5. L10:2084 weakly suppresses
  6. L11:102 weakly excites
  7. L6:1639 weakly suppresses
  8. L10:1205 weakly suppresses
  9. L8:2006 weakly suppresses
  10. L11:361 weakly suppresses
  11. L11:2324 weakly excites
  12. L11:2071 weakly suppresses
  13. L10:819 weakly excites
  14. L3:50 weakly excites
  15. L5:110 weakly suppresses
  16. L4:715 weakly suppresses
  17. L7:272 weakly excites
  18. L7:2174 weakly suppresses
  19. L2:493 weakly suppresses
  20. L11:1386 weakly suppresses
  21. L9:1313 weakly suppresses empty table cell separator bars
  22. L9:2413 weakly suppresses
  23. L11:623 weakly suppresses
  24. L7:2810 weakly excites

Feeds later layers

  1. L17:2075 excites
  2. L16:2529 excites
  3. L13:2162 excites
  4. L14:514 excites
  5. L17:202 excites
  6. L18:848 excites
  7. L15:2260 weakly excites
  8. L25:631 weakly suppresses
  9. L15:1797 weakly excites
  10. L13:1313 weakly suppresses
  11. L19:2749 weakly excites
  12. L17:603 weakly excites inside organisation and product names
  13. L15:1417 weakly excites
  14. L15:229 weakly excites
  15. L17:2518 weakly suppresses
  16. L16:662 weakly excites
  17. L13:1513 weakly suppresses
  18. L20:2149 weakly excites
  19. L17:375 weakly excites
  20. L25:1264 weakly suppresses
  21. L17:186 weakly excites
  22. L19:3055 weakly suppresses
  23. L21:598 weakly excites
  24. L17:1136 weakly suppresses

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.

1 values between -1.34 and -1.272 values between -0.72 and -0.651 values between -0.65 and -0.583 values between -0.58 and -0.5212 values between -0.52 and -0.4528 values between -0.45 and -0.3863 values between -0.38 and -0.31124 values between -0.31 and -0.24290 values between -0.24 and -0.17498 values between -0.17 and -0.10765 values between -0.10 and -0.03869 values between -0.03 and 0.04587 values between 0.04 and 0.10359 values between 0.10 and 0.17178 values between 0.17 and 0.2475 values between 0.24 and 0.3127 values between 0.31 and 0.3813 values between 0.38 and 0.454 values between 0.45 and 0.522 values between 0.52 and 0.591 values between 0.65 and 0.721 values between 0.72 and 0.791 values between 0.79 and 0.861 values between 0.86 and 0.931 values between 1.07 and 1.141 values between 1.14 and 1.201 values between 1.34 and 1.41offnegativepositiveoften

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. products
  2. tunnels
  3. john
  4. .biz
  5. Crush
  6. 苑
  7. .com
  8. (UInt
  9. forums
  10. balls

Suppresses

  1. ouver
  2. obot
  3. ormap
  4. ayar
  5. wer
  6. ivation
  7. ierten
  8. sát
  9. ierte
  10. UIManager