Neuron Explorer

Qwen3-0.6B

formal document and standards citations

A neuron in layer 22 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 lights up on the spot where a formal reference, standard number or document term is about to appear, like a standard code, a legal citation or a report title. When it was turned down, the model kept talking about documents, and when it was turned up the text drifted to unrelated topics.

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

    ,and I wrote this document.⏎QRcode is developed by DENSO Corporation. It is standardized by JIS and ISO.⏎This document is based on JIS X 0510(1999).⏎If you want to know

  2. #2 peak

    align with requirements such as HIPAA (Health Insurance Portability and Accountability Act) or GDPR (General Data Protection Regulation). This ensures

  3. #3 peak

    In addition, OSHA requires that SDS preparers provide specific minimum information as detailed in Appendix D of 29 CFR 1910.1200. The SDS prepar

  4. #4 peak

    law on June 15, 1976.⏎There are two main parts to what we call the HUD Code. The first part is called “Part 3280, Manufactured Home Construction and Safety Standards”. This concerns

  5. #5 peak

    18;15(2):57-61 DOI:http://dx.doi.org/10.4314/aas.v15i2.5⏎© 2018 Author. This work is

  6. #6 peak

    an SDS, the employer or designated person(s) should contact the manufacturer to obtain one.⏎Hazard Communication Standards⏎The Hazard Communication Standard (HCS) (29 CFR 1910.1200(g)), revised in

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: In the word "letter", the document is a word that is both a letter and a document. The word "letter" has 444444444444444444444

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

    Google Apps facilitates you to include a digital signature in header of the mails which are being sent from your domain, this will allow the receivers of

  2. #2 peak

    know you are in danger and call the police.⏎• Open your own savings account to increase your independence.⏎• Leave money, an extra set of keys, copies of important documents, and extra clothes with someone you trust.⏎DURING A VIO

  3. #3 peak

    aaight. As the dictionary indicates, aight’s colloquial English, so you should probably avoid dropping it into your college essays, newspaper columns, and legal documents. It looks to be the most recent evolution of the “all right”

  4. #4 peak

    who came to the United States before us, the people in our country who are still here, and the<|endoftext|>What Is a Warning Letter?⏎A warning letter is a letter addressed to an erring individual whether in a company of any industry, a

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. L10:1724 suppresses
  2. L6:2472 excites
  3. L10:1881 excites
  4. L8:1609 suppresses
  5. L11:626 excites
  6. L20:139 suppresses
  7. L20:1875 weakly suppresses
  8. L21:1654 weakly excites
  9. L13:1585 weakly excites
  10. L17:1646 weakly suppresses
  11. L21:522 weakly excites
  12. L20:1477 weakly suppresses
  13. L9:574 weakly excites
  14. L8:1274 weakly suppresses
  15. L10:1248 weakly excites
  16. L20:443 weakly excites
  17. L4:2152 weakly suppresses
  18. L21:2545 weakly suppresses
  19. L17:133 weakly suppresses
  20. L6:2789 weakly suppresses
  21. L7:2244 weakly suppresses
  22. L9:1785 weakly suppresses
  23. L6:1409 weakly suppresses
  24. L7:2480 weakly excites

Feeds later layers

  1. L27:1502 suppresses
  2. L24:575 suppresses
  3. L26:1107 suppresses
  4. L25:2989 suppresses
  5. L23:1372 suppresses
  6. L26:2985 suppresses
  7. L26:1878 suppresses
  8. L25:1028 suppresses
  9. L26:2658 suppresses
  10. L26:1350 suppresses
  11. L27:1109 suppresses
  12. L26:2876 suppresses
  13. L24:2330 suppresses
  14. L25:847 weakly suppresses
  15. L25:2969 weakly suppresses
  16. L23:1823 weakly suppresses
  17. L25:2765 weakly suppresses
  18. L27:2421 weakly suppresses
  19. L24:2980 weakly suppresses the word 'page' in references
  20. L27:1643 weakly suppresses
  21. L27:2984 weakly suppresses
  22. L27:560 weakly suppresses
  23. L23:2622 weakly suppresses
  24. L26:1811 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 -16.05 and -15.441 values between -13.02 and -12.421 values between -12.42 and -11.811 values between -11.81 and -11.211 values between -9.39 and -8.781 values between -8.78 and -8.181 values between -8.18 and -7.571 values between -6.97 and -6.362 values between -6.36 and -5.763 values between -5.76 and -5.157 values between -5.15 and -4.558 values between -4.55 and -3.943 values between -3.94 and -3.3411 values between -3.34 and -2.7320 values between -2.73 and -2.1329 values between -2.13 and -1.5248 values between -1.52 and -0.92292 values between -0.92 and -0.312444 values between -0.31 and 0.30941 values between 0.30 and 0.9072 values between 0.90 and 1.517 values between 1.51 and 2.112 values between 2.11 and 2.724 values between 2.72 and 3.322 values between 3.32 and 3.931 values between 3.93 and 4.531 values between 5.14 and 5.741 values between 6.35 and 6.951 values between 6.95 and 7.561 values between 7.56 and 8.16offnegativepositiveoften

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. uhe
  2. berg
  3. onest
  4. 贞
  5. icolor
  6. rtrim
  7. رام
  8. urious
  9. URY
  10. ridge

Suppresses

  1. document
  2. 文档
  3. Document
  4. documents
  5. Document
  6. <Document
  7. document
  8. documents
  9. documento
  10. PDF