Neuron Explorer

Qwen3-0.6B

'in' in 'keep in mind'

A neuron in layer 25 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 'in' in phrases like 'bear in mind' and 'keep in mind', where 'in' comes right before 'mind'. Turning it down made the model repeat the word 'in' over and over, which shows how strongly it is tied to this one small word.

Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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

    them fascinating. After you have mastered some fundamental activities, you’ll be able to try new ones and incorporate a brand new twist for the previous kinds. And bear in mind, it’s in no way too early to start creating new memories. >>>

  2. #2 peak

    forest for one use if it's going to compromise other uses? Our forests are incredibly important for harvesting but also for tourism [and] we want to always bear in mind the possible tradeoffs.⏎Tourism is a huge asset to Maine

  3. #3 peak

    Margo Wood|⏎|Number Of Pages||408|⏎Write a review⏎Note: HTML is not translated!⏎Rating: Bad Good⏎Enter the code in the box below:<|endoftext|>Everyday researchers around the world look for new therapies in

  4. #4 peak

    specifically occurred in the older days when the trading was done via wind-powered ships. Nowadays, this scenario has become obsolete.⏎Below mentioned are the facts to be kept in mind regarding re export:⏎- The state of items

  5. #5 peak

    . As a result, they are said to bring relief from illnesses including heart problems, asthma, skin disorder, and arthritis.⏎In the end, you must keep in mind that Ayurvedic<|endoftext|>The number of women who

  6. #6 peak

    14 x 17 sheet of paper, using the same compressed charcoal. This way you will not get too caught up in detail. Keep these points in mind: Find the basic, simple shapes to describe the character. Find distinctive features

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: in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in in

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

    by behaviors including argumentation and silence. These behaviors, in turn, function to reinstate white racial equilibrium and prevent any meaningful cross-racial dialogue. In this in-depth exploration, DiAngelo examines how white fragility develops, how it protects

  2. #2 peak

    practices⏎This past year, I have had rewarding opportunities to observe traditional food cultures in varied regions of the world. These are:⏎Athabascan Indian in the interior of Alaska (the traditional Tanana Chiefs Conference tribal lands) in July,

  3. #3 peak

    on either side, down to a massive bend in the Urubamba River. What could have motivated the Incas to undertake such a construction at this remote location in the Andean cordillera? Machu Picchu, built over 7

  4. #4 peak

    concomitant users of warfarin and a skeletal muscle relaxant and, as pre-specified negative control comparators, among concomitant users of an in

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. L21:1935 strongly suppresses
  2. L10:262 strongly suppresses
  3. L17:706 excites
  4. L18:2356 excites
  5. L4:289 excites
  6. L18:483 excites
  7. L18:869 suppresses
  8. L22:535 excites
  9. L16:1557 excites 'in' meaning 'in this case' or 'in between'
  10. L16:1807 suppresses 'in' after a motion verb
  11. L8:351 suppresses 'in' after 'high' or 'dressed'
  12. L5:490 excites
  13. L20:1574 excites
  14. L23:2829 excites
  15. L3:493 suppresses
  16. L19:2592 suppresses
  17. L14:1625 excites the word 'in' after a noun
  18. L19:704 excites
  19. L4:1966 suppresses
  20. L21:1269 suppresses "the" in "on the other hand"
  21. L16:2120 suppresses
  22. L14:1122 suppresses
  23. L20:2850 excites
  24. L8:3068 suppresses

Feeds later layers

  1. L26:207 strongly excites
  2. L27:405 suppresses
  3. L26:825 suppresses
  4. L26:584 suppresses
  5. L27:1621 suppresses
  6. L27:2429 weakly suppresses
  7. L26:786 weakly excites
  8. L27:651 weakly suppresses
  9. L27:2594 weakly suppresses
  10. L27:2982 weakly suppresses
  11. L26:2107 weakly excites
  12. L26:164 weakly excites
  13. L26:1662 weakly excites
  14. L26:938 weakly excites
  15. L27:142 weakly suppresses
  16. L26:186 weakly excites
  17. L27:1426 weakly suppresses
  18. L27:1795 weakly excites
  19. L26:2401 weakly suppresses
  20. L27:2277 weakly suppresses
  21. L27:2506 weakly suppresses
  22. L26:82 weakly suppresses
  23. L27:1522 weakly excites
  24. L27:2486 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 -27.58 and -26.481 values between -25.39 and -24.292 values between -17.72 and -16.621 values between -16.62 and -15.535 values between -15.53 and -14.434 values between -14.43 and -13.332 values between -13.33 and -12.244 values between -12.24 and -11.144 values between -11.14 and -10.054 values between -10.05 and -8.954 values between -8.95 and -7.868 values between -7.86 and -6.7614 values between -6.76 and -5.6618 values between -5.66 and -4.5737 values between -4.57 and -3.4751 values between -3.47 and -2.38106 values between -2.38 and -1.28702 values between -1.28 and -0.192807 values between -0.19 and 0.9178 values between 0.91 and 2.0117 values between 2.01 and 3.109 values between 3.10 and 4.209 values between 4.20 and 5.296 values between 5.29 and 6.393 values between 6.39 and 7.484 values between 7.48 and 8.582 values between 8.58 and 9.681 values between 10.77 and 11.872 values between 12.96 and 14.062 values between 15.15 and 16.25offnegativepositiveoften

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. ient
  2. ことです
  3. ients
  4. 投标人
  5. @stop
  6. ilight
  7. onto
  8. Stanton
  9. görüş
  10. lsen

Suppresses

  1. in
  2. sense
  3. 日在
  4. context
  5. context
  6. 在
  7. 而在
  8. 就是在
  9. 是在
  10. 在这种