Neuron Explorer

Qwen3-0.6B

'the' after 'of' and similar words

A neuron in layer 17 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 'the' when it follows words like 'of', 'combination of' or 'without', which often start a noun phrase. The pattern is loose, and turning it up gave repetitive, unfocused text.

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

    for the construction of skyscrapers, bridges, and other structures, dramatically reducing carbon dioxide. Every ton of cement manufactured for use in concrete emits a ton of the greenhouse gas carbon dioxide (CO2) into the atmosphere. Worldwide, cement production

  2. #2 peak

    an introduction by the editor including the latest research, references, and updates in the field. This book is enhanced with numerous color illustrations, a comprehensive index, and the many computer program examples encourage reader

  3. #3 peak

    wrinkles, crepey skin, or razor bumps, a board certified dermatologist or dermasurgeon may help you by offering a single treatment or any combination of the following:⏎Cosmetic dermatology - Non-surgical procedures⏎- Laser treatments⏎

  4. #4 peak

    , too many doctors still do not think of endometriosis when their patients report symptoms such as intermittent constipation or diarrhoea, or alternating bouts of the two or bleeding from the rectum.⏎Most importantly, they do not think to

  5. #5 peak

    , which is caused by a recently revealed coronavirus. Those who are infected with the virus will experience mild to modest respiratory sickness. They can recover from the illness without the need for special treatment. The older population is more vulnerable to the virus. People

  6. #6 peak

    a single speck of dust in the eye.⏎If you haven’t used contact lenses to correct your vision, drawing blood several times a day might still sound like the preferred option. But contact lenses have been used as medical devices for hundreds of years

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: and of the or of of of of of of of of of of of of of of of of of of of of of (2,4,6, of or of of of of of of of of of of of 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

    oll) martix. It is an extension of Bullseye matrix problem.⏎Who has power to do everything in this world?⏎Close MATLAB with keyboard⏎Find the treasures in MATLAB Central and discover how the community can help you!⏎Choose a web

  2. #2 peak

    early period, 1853–1930⏎Our editors will review what you’ve submitted and determine whether to revise the article.Join Britannica's Publishing Partner Program and our community of experts to gain a global audience for your work

  3. #3 peak

    in-person schooling, open parks, and playgrounds will return. But, we do have a sense of what we can—and must—do in this moment, this liminal time.<|endoftext|>When we were in kindergarten and playing sports and our team

  4. #4 peak

    the only ones using baby powder. Many adults continue to use baby powder as part of their daily ritual in place of deodorant or for feminine hygiene. Remember the Shower

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. L16:2248 suppresses
  2. L10:1377 suppresses
  3. L16:792 excites "the" inside a prepositional phrase
  4. L9:24 excites "the" before a named organisation or title
  5. L4:52 excites
  6. L11:331 excites
  7. L11:875 suppresses
  8. L4:73 suppresses
  9. L10:120 suppresses
  10. L9:662 suppresses
  11. L16:2062 excites "other" and contrast words
  12. L11:1168 excites
  13. L16:419 excites
  14. L13:1234 suppresses
  15. L12:435 suppresses
  16. L8:636 suppresses
  17. L6:61 suppresses
  18. L5:225 excites
  19. L2:205 excites
  20. L16:1851 suppresses
  21. L15:1848 suppresses
  22. L16:1309 excites
  23. L10:1162 excites
  24. L16:1212 suppresses

Feeds later layers

  1. L20:457 excites
  2. L19:1550 suppresses
  3. L18:314 excites
  4. L21:592 suppresses
  5. L18:884 weakly suppresses
  6. L18:1477 weakly suppresses
  7. L20:932 weakly excites 'of' after a count or quantity
  8. L18:3050 weakly suppresses
  9. L19:684 weakly suppresses
  10. L22:260 weakly suppresses
  11. L20:901 weakly suppresses
  12. L18:2513 weakly excites
  13. L22:360 weakly excites
  14. L20:2025 weakly suppresses
  15. L18:217 weakly excites
  16. L20:880 weakly suppresses
  17. L19:769 weakly suppresses
  18. L27:1992 weakly suppresses
  19. L19:1475 weakly suppresses
  20. L18:2842 weakly excites
  21. L18:2523 weakly excites 'the' after 'is' or 'was' before a role
  22. L20:64 weakly suppresses
  23. L21:1579 weakly excites
  24. L18:1076 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 -7.96 and -7.451 values between -6.42 and -5.903 values between -5.39 and -4.881 values between -4.88 and -4.364 values between -4.36 and -3.856 values between -3.85 and -3.335 values between -3.33 and -2.8211 values between -2.82 and -2.3015 values between -2.30 and -1.7916 values between -1.79 and -1.2830 values between -1.28 and -0.76529 values between -0.76 and -0.252913 values between -0.25 and 0.27273 values between 0.27 and 0.7821 values between 0.78 and 1.3022 values between 1.30 and 1.8114 values between 1.81 and 2.3211 values between 2.32 and 2.8411 values between 2.84 and 3.355 values between 3.35 and 3.873 values between 3.87 and 4.385 values between 4.38 and 4.902 values between 4.90 and 5.411 values between 5.41 and 5.921 values between 5.92 and 6.441 values between 6.95 and 7.471 values between 7.47 and 7.981 values between 9.01 and 9.521 values between 12.10 and 12.61offnegativepositiveoften

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. positor
  2. izons
  3. icias
  4. s
  5. earn
  6. pires
  7. #__
  8. íc
  9. 变现
  10. ibir

Suppresses

  1. competent
  2. PTS
  3. inality
  4. heuristic
  5. weg
  6. 없이
  7. 醒
  8. seule
  9. قة
  10. فة