Neuron Explorer

Qwen3-0.6B

"before" after a do-this-first advice clause

A neuron in layer 9 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 fires on the word "before" (or "prior to" or "to") when it comes after advice such as "talk to your doctor before starting" or "wait until the offer is accepted before acting". Turning it up or down did not clearly change the topic of the generated text.

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

    their first name.⏎If you offer assistance, wait until the offer is accepted before acting. Then listen to or ask for instructions.⏎Disability: related terms and

  2. #2 peak

    you think you may have a medical emergency, call your doctor or (in the United States) 911 immediately. Always seek the advice of your doctor before starting or changing treatment. Medical professionals who provide responses to health-related

  3. #3 peak

    very disciplined and had a strong work ethic. He avoided any possible distractions such as alcohol and drugs.⏎Rockefeller was patient and waited until he had the perfect opportunity before investing in a project.⏎6. He was also very shrewd, making

  4. #4 peak

    AIN AFFILIATE LINKS*⏎Disclaimer: I am not a medical doctor and nothing in this article should be taken as medical advice. Please talk to your doctor before using any of the herbs and/or remedies mentioned in this article. Yarrow is

  5. #5 peak

    I might discuss the notion of a cell and where it came from. Sometimes I do a bit of review if I think it’s essential that students know certain things before we start the lecture. During this introduction time, I always write the topics for

  6. #6 peak

    ARE PROVIDER IMMEDIATELY IF YOU THINK YOU MAY HAVE A MEDICAL EMERGENCY. Always seek the advice of your physician or other qualified health provider prior to starting any new treatment or with any questions you may have regarding a medical condition.

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: What is the role of the human brain in the modern world? Answer: The human brain in the modern world is made up of a lot of small parts? This is the brain? Or is it the part of the human brain that is

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

    the property, it is considered real property.⏎Adaptability: If an asset is purchased and adapted specifically to the property

  2. #2 peak

    and Îles de la Madeleine), Nova Scotia (including present-day N.B. and P.E.I.). Hudson's Bay Co. still administers Rupert's Land. Louisiana is ceded to Spain by France.⏎Following the Constitutional Act, Quebec

  3. #3 peak

    needing the author's permission or paying licensing fees and royalties. Section 107 of the Copyright Act of 1976 specifically provides for “fair use” of copyrighted materials for teaching, scholarship, or research, as well as for

  4. #4 peak

    .S.⏎Influential Women: U. S.⏎Influential Women: World⏎Primary Sources: U.S.⏎Primary Sources: World⏎Public Domain and Copyright Information⏎Large

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. L8:1477 excites
  2. L8:2704 suppresses
  3. L7:1850 weakly excites
  4. L7:1173 weakly suppresses
  5. L8:2485 weakly suppresses
  6. L7:1674 weakly suppresses
  7. L6:2855 weakly suppresses
  8. L5:1915 weakly excites
  9. L3:2686 weakly suppresses
  10. L6:2260 weakly suppresses
  11. L8:1522 weakly excites
  12. L3:2457 weakly suppresses
  13. L8:140 weakly excites
  14. L1:373 weakly excites
  15. L4:843 weakly excites
  16. L4:1746 weakly excites
  17. L8:2415 weakly suppresses words after 'will be' or 'subject to' in formal text
  18. L5:96 weakly suppresses
  19. L5:1868 weakly excites
  20. L5:1854 weakly suppresses
  21. L8:1146 weakly excites
  22. L8:3066 weakly excites
  23. L7:1319 weakly excites
  24. L6:2844 weakly suppresses

Feeds later layers

  1. L18:2080 excites
  2. L10:2338 suppresses
  3. L12:2895 weakly excites
  4. L21:1260 weakly excites
  5. L10:1570 weakly excites
  6. L11:1569 weakly excites
  7. L10:2459 weakly suppresses
  8. L12:2375 weakly excites
  9. L11:1370 weakly excites 'then' in step-by-step sequences
  10. L16:2732 weakly excites
  11. L14:457 weakly excites
  12. L16:2862 weakly suppresses
  13. L11:2935 weakly excites
  14. L16:2893 weakly suppresses
  15. L16:2872 weakly excites
  16. L10:2120 weakly excites
  17. L12:2720 weakly excites
  18. L11:2649 weakly excites
  19. L11:462 weakly excites
  20. L20:419 weakly suppresses
  21. L11:2917 weakly suppresses
  22. L11:1201 weakly excites
  23. L21:2063 weakly suppresses
  24. L12:1953 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.

1 values between -2.59 and -2.441 values between -1.35 and -1.191 values between -1.19 and -1.034 values between -0.88 and -0.726 values between -0.72 and -0.5733 values between -0.57 and -0.41201 values between -0.41 and -0.25696 values between -0.25 and -0.101678 values between -0.10 and 0.06955 values between 0.06 and 0.21255 values between 0.21 and 0.3750 values between 0.37 and 0.5315 values between 0.53 and 0.685 values between 0.68 and 0.841 values between 0.84 and 0.991 values between 0.99 and 1.151 values between 1.15 and 1.311 values between 1.31 and 1.461 values between 1.62 and 1.771 values between 1.93 and 2.091 values between 3.49 and 3.65offnegativepositiveoften

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. icias
  2. arus
  3. holland
  4. edriver
  5. GINE
  6. crian
  7. _DER
  8. ⾏
  9. asics
  10. derivatives

Suppresses

  1. 皱
  2. bedo
  3. Loop
  4. tav
  5. ",⏎⏎
  6. bibli
  7. %⏎⏎
  8. _frequency
  9. Loop
  10. imeType