Neuron Explorer

Qwen3-0.6B

unclear: formal and legal terms

A neuron in layer 4 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 responds to the later part of a longer formal or technical word, such as 'ations' in appellations or 'pher' in Shepherd. It is hard to say what unifies the examples, and turning it up or down did not change the writing in a clear way.

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

    all AVA’s are appellations, but not all appellations are AVA’s. For example, a state or county name may be used

  2. #2 peak

    , the United States Code or the Arizona Revised Statutes). The statutes that are published in a state’s code are grouped by subject matter into Titles, as in Title 11 of the United States Code (bank

  3. #3 peak

    while solving a problem, how it makes decisions and how it works.⏎The government Support:⏎A committee has been formed under the chairmanship of NITI Aayog Vice Chairman Rajiv Kumar to outline the Artificial

  4. #4 peak

    assumes any responsibility for errors, omissions or contrary interpretation of the subject matter. Walmart does not sponsor, recommend or endorse any third party, product, service or information provided on this site.⏎If you have dietary

  5. #5 peak

    complications that we will talk about shortly.⏎Below is a list of common health challenges that being overweight can cause your German Shepherd⏎- Arthritis: As German Shepherds age, they naturally become prone to hip dysplasia and arthritis; however

  6. #6 peak

    the<|endoftext|>Obese & Fat German Shepherd⏎Like humans, dogs can become obese for several reasons, and the effect is just as severe in them. German Shepherds are prone to becoming obese, and in this article, you will find out

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: the person who is not in the place of the money, and the person who is not in the place of the goods, and the person who is not in the place of the service, and the person who is not in the place of

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

    ⏎Let’s break that down a bit. The first important concept is the part about “a course of vexatious conduct”. In this case, vexatious means distressing or troubled. A course of behaviour means that the behaviour occurs more

  2. #2 peak

    . If you or a loved one have already been diagnosed with a head injury, and you now face disability, mental impairment, or worse you should contact a personal injury attorney who can advise you because you may be entitled to a settlement.⏎The most

  3. #3 peak

    acquire enough for domicile of choice.⏎I. Traditional Approaches⏎i. Traditional approach: law of where the injury occurs.⏎ii. Non-intentional Torts⏎1. Alabama Great Southern RR v. Carroll (AL, 18

  4. #4 peak

    us accidents are caused by driver error, several other factors can play a role in tragic accidents, including⏎- Failure to maintain buses by school districts⏎- Negligent hiring of bus drivers by school districts⏎- School buses with defective parts that

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. L3:1101 suppresses
  2. L3:85 excites
  3. L0:2257 suppresses
  4. L3:1146 suppresses
  5. L3:2205 excites
  6. L3:2369 weakly suppresses
  7. L1:2801 weakly excites
  8. L3:813 weakly excites
  9. L3:115 weakly suppresses
  10. L0:1236 weakly suppresses
  11. L0:2815 weakly suppresses
  12. L3:691 weakly excites
  13. L1:1770 weakly suppresses
  14. L3:153 weakly excites
  15. L2:1710 weakly suppresses
  16. L0:2517 weakly excites
  17. L1:139 weakly excites
  18. L1:2175 weakly suppresses
  19. L3:271 weakly excites
  20. L0:529 weakly excites
  21. L1:2877 weakly excites
  22. L3:66 weakly excites
  23. L3:1749 weakly excites
  24. L1:1322 weakly excites

Feeds later layers

  1. L19:642 weakly suppresses
  2. L21:775 weakly excites the article 'a' before a noun
  3. L7:793 weakly suppresses
  4. L7:846 weakly suppresses
  5. L20:2802 weakly excites
  6. L19:3026 weakly excites
  7. L21:2410 weakly excites
  8. L9:1970 weakly excites
  9. L18:235 weakly suppresses
  10. L18:2407 weakly excites
  11. L10:2172 weakly excites
  12. L9:593 weakly suppresses
  13. L7:1938 weakly suppresses
  14. L5:435 weakly suppresses
  15. L22:808 weakly suppresses
  16. L18:193 weakly suppresses
  17. L13:2661 weakly suppresses
  18. L9:2540 weakly excites
  19. L8:2308 weakly suppresses
  20. L16:2160 weakly excites
  21. L5:3038 weakly suppresses
  22. L9:1401 weakly excites
  23. L14:187 weakly excites
  24. L8:318 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 -1.84 and -1.781 values between -1.78 and -1.723 values between -0.67 and -0.611 values between -0.61 and -0.543 values between -0.54 and -0.485 values between -0.48 and -0.4215 values between -0.42 and -0.3630 values between -0.36 and -0.3063 values between -0.30 and -0.23157 values between -0.23 and -0.17271 values between -0.17 and -0.11526 values between -0.11 and -0.05809 values between -0.05 and 0.01749 values between 0.01 and 0.08539 values between 0.08 and 0.14361 values between 0.14 and 0.20200 values between 0.20 and 0.2691 values between 0.26 and 0.3247 values between 0.32 and 0.3923 values between 0.39 and 0.458 values between 0.45 and 0.514 values between 0.51 and 0.571 values between 0.57 and 0.63offnegativepositiveoften

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. rag
  2. Star
  3. .star
  4. 袤
  5. vard
  6. 为企业
  7. 航空航天
  8. 是一部
  9. star
  10. 过得

Suppresses

  1. erral
  2. ՚
  3. encion
  4. \xaa
  5. elight
  6. 罍
  7. abilidade
  8. tối
  9. edith
  10. ustria