Neuron Explorer

Qwen3-0.6B

"of" before a field or department name

A neuron in layer 14 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 small word right after a job title or role, such as "professor of", "chair of" or "Director of", just before the subject area is named. It also fires on "no matter what". When pushed up, the written text drifted into loose, generic sentences, which does not clearly match the trigger.

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

    <|endoftext|>Relationship Between Diabetes and Mental Health⏎"I see a lot of diabetes in my clients." exclaims Dr. William H. Wilson, Professor of Psychiatry and Director of Inpatient Psychiatric Services at Oregon Health & Science University.⏎A simple statement that

  2. #2 peak

    tackle, and our legislators and community leaders should rise to the task of protecting all citizens equally, no matter what job you choose or where you choose to dine. We all know where this pollut

  3. #3 peak

    . When we prove their role in the aging process, then we can work toward breaking them.⏎Hope for Sore Joints⏎Robert Spilker, chair of biomedical engineering at Rensselaer<|endoftext|>A probe flying through deep space has

  4. #4 peak

    times during his candidacy, and always put on a brave face no matter what the circumstances were, and did what was right no matter what!⏎George Washington was indispensable in launching the U.S government. Although many opposed his

  5. #5 peak

    : Pure and Applied Optics, published by the Institute of Physics in London.⏎Giulio Fanti, a professor of mechanical engineering at the University of Padua, and his colleague, Roberto Maggi

  6. #6 peak

    World Immunization Week, read the personal stories of several key advocacy partners, including the American Academy of Pediatrics! Emory University Associate Professor of Pediatrics in the Division of Pediatric Infectious Diseases, Rana Chakr

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: I think I (the) [1] [1. 1. 1. 1. 1. [1. 1. 1 1 1 2 1

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

    Bible Battles take place in a relatively small area of the Levant (which, in itself, is only a

  2. #2 peak

    need to print out the deck of picture cards and at least one board for each player. Each player will also need a dozen or so dried beans or pennies to mark their board.⏎Each player gets a game board (which shows 16

  3. #3 peak

    of morn;⏎Loose the hounds and let them go,⏎Wax the cord and bend the bow."⏎SO long as the new moon returns in heaven a bent, beautiful bow, so long will the fascination of archery keep hold of

  4. #4 peak

    lots of traditional publishers choose to publish their books with a huge publishing house.⏎Many authors that select to self-publish their books locate typical publishers and also agents to be a valuable resource of details concerning their story

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. L2:70 suppresses
  2. L2:75 excites
  3. L13:999 excites
  4. L1:951 suppresses
  5. L3:421 excites
  6. L1:1464 suppresses
  7. L2:549 excites
  8. L3:59 weakly suppresses
  9. L13:596 weakly excites
  10. L1:1029 weakly suppresses
  11. L11:204 weakly excites
  12. L6:16 weakly suppresses
  13. L2:49 weakly excites
  14. L1:146 weakly suppresses
  15. L4:245 weakly suppresses
  16. L0:1301 weakly excites
  17. L0:1224 weakly excites
  18. L0:1 weakly excites
  19. L8:10 weakly suppresses
  20. L2:483 weakly excites
  21. L0:178 weakly excites
  22. L1:2138 weakly excites
  23. L1:2683 weakly suppresses
  24. L1:2801 weakly suppresses

Feeds later layers

  1. L16:2843 suppresses
  2. L17:967 suppresses
  3. L16:1891 suppresses
  4. L16:469 excites
  5. L17:2272 suppresses
  6. L16:266 excites
  7. L16:1957 suppresses
  8. L17:279 excites
  9. L15:23 weakly excites
  10. L16:23 weakly excites
  11. L15:43 weakly excites
  12. L16:549 weakly excites
  13. L18:25 weakly excites
  14. L16:166 weakly excites
  15. L18:27 weakly excites
  16. L18:2307 weakly suppresses
  17. L18:16 weakly excites
  18. L19:26 weakly excites
  19. L19:80 weakly excites
  20. L21:1499 weakly suppresses question marks ending a question
  21. L16:10 weakly excites
  22. L16:15 weakly excites
  23. L16:61 weakly excites
  24. L16:7 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 -4.01 and -3.802 values between -3.37 and -3.164 values between -2.95 and -2.7417 values between -2.74 and -2.5216 values between -2.52 and -2.3118 values between -2.31 and -2.1044 values between -2.10 and -1.8961 values between -1.89 and -1.6776 values between -1.67 and -1.4692 values between -1.46 and -1.25158 values between -1.25 and -1.04203 values between -1.04 and -0.82264 values between -0.82 and -0.61248 values between -0.61 and -0.40278 values between -0.40 and -0.19307 values between -0.19 and 0.03324 values between 0.03 and 0.24307 values between 0.24 and 0.45269 values between 0.45 and 0.67238 values between 0.67 and 0.88212 values between 0.88 and 1.09204 values between 1.09 and 1.30142 values between 1.30 and 1.52110 values between 1.52 and 1.73100 values between 1.73 and 1.9460 values between 1.94 and 2.1546 values between 2.15 and 2.3730 values between 2.37 and 2.5841 values between 2.58 and 2.7917 values between 2.79 and 3.009 values between 3.00 and 3.226 values between 3.22 and 3.431 values between 3.43 and 3.642 values between 3.64 and 3.851 values between 4.28 and 4.49offnegativepositiveoften

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. {}).
  2. identally
  3. activo
  4. @Service
  5. ycles
  6. ).⏎
  7. 的就是
  8. algu
  9. mars
  10. ,)⏎

Suppresses

  1. itia
  2. zier
  3. athering
  4. "},{"
  5. iev
  6. ","
  7. AKER
  8. ','
  9. overnment
  10. Claud