Neuron Explorer

Qwen3-0.6B

'of' after 'those' or 'these'

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 responds to the word 'of' when it comes after 'those' or 'these', as in 'for those of you' or 'these of us'. Turning it up did not produce text that clearly shows 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

    roofs. Green walls emerged as clear winners<|endoftext|>Have you ever found yourself wishing that the list you just typed in MS Word was alphabetized? Or, for those of you who are good with Excel, you're probably wishing you typed the list in

  2. #2 peak

    visual for teachers. I just came across it while I was going through my Google Plus feeds and thought you may want to have a look as well. For those of you not yet familiar with what SAMR model is all about , here is a

  3. #3 peak

    physically disabled.⏎This is what separates autism from other disabilities.⏎In my presentation, I’m going to be using the word “neurotypical.” For those of you who don’t know what that means, it is a term in the autism

  4. #4 peak

    , given enough time and sufficient access, they hope to catch a glimpse of the famous ship too. Only time will tell if that will actually happen, but those of us who are fans of Shackleton and his story are

  5. #5 peak

    of all, soil pudding tastes great with ingredients like Oreos, chocolate pudding, and gummy worms. Click here to view a recipe for edible soil.⏎Those of us who have ever had to teach the parts of the cell know how tricky it

  6. #6 peak

    Guide: Where can find journal ...⏎Functioning as a research and publishing network, ScienceOpen offers open access to more than 28 million articles in all areas of science. Although you do need to register to view the full text of the articles

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: if I have 3 numbers that I have 3 numbers that I have three numbers that I have 3 numbers that I have 3 numbers that I have 3 numbers - 3 numbers - 3 - - - - - -

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 schools shown by ward or type. Selecting a school will reveal detailed information including the budget line items that the funding supported.⏎In order to realize their potential as a tool for addressing achievement

  2. #2 peak

    water. Repeat if necessary.⏎Your body makes hydrogen peroxide to fight infection which must be present for our immune system to function correctly. White blood cells are known as Leukocytes. A sub-class of Leukocytes called Neutrophils

  3. #3 peak

    do their part to protect Americans, food safety can’t end with the FDA. It must continue all the way to our dinner tables, which means that we, as consumers, need to know and practice proper food safety habits in our everyday lives.

  4. #4 peak

    the lives of about 7.6 million people in the world. Physicians and researchers who specialize in the study, diagnosis, treatment, and prevention of cancer are called oncologists.⏎What causes cancer?⏎Cancer is ultimately the result of cells 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. L16:1317 strongly excites
  2. L4:298 strongly excites
  3. L10:661 strongly suppresses
  4. L8:253 strongly suppresses
  5. L2:149 strongly suppresses
  6. L3:654 strongly excites article 'a' after 'here are'
  7. L16:1418 strongly excites
  8. L16:341 strongly suppresses
  9. L11:282 strongly suppresses
  10. L9:1062 strongly excites
  11. L7:146 strongly excites
  12. L9:55 strongly excites
  13. L2:1426 strongly excites
  14. L12:642 strongly excites
  15. L5:478 strongly excites
  16. L4:450 strongly excites
  17. L7:327 strongly excites
  18. L10:1190 strongly suppresses
  19. L2:326 strongly suppresses
  20. L6:424 strongly excites
  21. L8:1220 strongly suppresses
  22. L5:13 strongly excites
  23. L2:502 strongly excites 'other' after a preposition
  24. L15:2146 excites

Feeds later layers

  1. L24:434 strongly suppresses
  2. L20:1047 strongly suppresses
  3. L19:654 strongly excites
  4. L19:626 strongly suppresses
  5. L19:2522 strongly suppresses
  6. L20:807 strongly suppresses
  7. L27:60 strongly excites
  8. L25:812 strongly suppresses
  9. L25:377 strongly suppresses
  10. L25:335 strongly suppresses
  11. L26:111 strongly suppresses
  12. L19:2593 suppresses
  13. L24:109 excites
  14. L20:240 excites
  15. L25:2260 suppresses
  16. L24:419 suppresses
  17. L24:488 suppresses
  18. L24:360 excites
  19. L24:443 suppresses
  20. L27:969 suppresses
  21. L19:254 excites
  22. L26:246 suppresses
  23. L20:1989 excites
  24. L27:477 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 -8.91 and -8.471 values between -7.59 and -7.151 values between -5.82 and -5.382 values between -5.38 and -4.943 values between -4.49 and -4.052 values between -4.05 and -3.614 values between -3.61 and -3.176 values between -3.17 and -2.7316 values between -2.73 and -2.2810 values between -2.28 and -1.8431 values between -1.84 and -1.4058 values between -1.40 and -0.96143 values between -0.96 and -0.52854 values between -0.52 and -0.072325 values between -0.07 and 0.37235 values between 0.37 and 0.8184 values between 0.81 and 1.2548 values between 1.25 and 1.6923 values between 1.69 and 2.1421 values between 2.14 and 2.5812 values between 2.58 and 3.0210 values between 3.02 and 3.464 values between 3.46 and 3.905 values between 3.90 and 4.353 values between 4.35 and 4.791 values between 4.79 and 5.232 values between 5.23 and 5.672 values between 7.88 and 8.321 values between 8.32 and 8.77offnegativepositiveoften

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. 的一个
  2. 还有一个
  3. 一个
  4. 有一个
  5. 了一个
  6. człowiek
  7. ʏ
  8. quette
  9. 一个问题
  10. 的一款

Suppresses

  1. asures
  2. elines
  3. adients
  4. евых
  5. ницы
  6. inations
  7. plies
  8. akers
  9. cepts
  10. ersions