Neuron Explorer

Qwen3-0.6B

'in' or 'on' ending a phrase

A neuron in layer 13 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 words 'in' or 'on' that come right after a verb or noun and finish a phrase, such as 'hold in', 'moisture in' or 'grid snap on'. Turning it up made the model produce strings of 'and' or 'the' with no meaning.

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

    ⏎Whether it be driving the kids to extracurriculars, getting up at ungodly hours just to walk the dog or get a quick gym session in, or the daily grind of our careers; interdental cleaning such as f

  2. #2 peak

    you can change the shape of the curve on the spline.⏎Use the same settings as stated in the Simple Mesh tutorial (16 Grid, Grid snap on).⏎Lets start off by creating a circle with a radius of 64 (

  3. #3 peak

    in improving your health and or the way you look.⏎Calorie Intake and You⏎We’ll be looking into the idea of energy balance ie:calories in versus calories out. Today we’ll take a look out the calories you expend-energy

  4. #4 peak

    There were shoals and sand everywhere, blocking nine-tenths of the land altogether, and doing their best to block the other tenth where the great rivers run in.⏎The North Sea coast of the German Empire was the most heavily fortified stretch of

  5. #5 peak

    Namibia, a former diamond mining town whose population steadily declined after WWI, when diamond prices crashed, and Balestrino, Italy, whose inhabitants were moved in 1953 after several earthquakes brought them to the conclusion that the area

  6. #6 peak

    eter⏎- Kidney stones⏎- Suppressed immune system⏎- Immobility for a long period⏎- Holding your pee: If you tend to hold in your pee during the day or not empty your bladder completely when you do go to

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 first time I saw the article "Dish 1011112221122230111201112011120111221

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

    Israel⏎- *Corresponding Author:⏎- Yeongha Wook⏎Department of Pediatrics,⏎Soroka University Medical Center,⏎E-mail: [email protected]⏎Received: 01-Nov-2022, Manuscript No

  2. #2 peak

    preprocessors are given. It executes them in the following order:⏎removes instances with unknown target values⏎removes empty columns⏎imputes missing

  3. #3 peak

    2, and we know that,⏎Lines with the same slope are parallel.⏎Hence, the given lines are parallel.⏎Let's consider another set of equations.⏎y = 2x +3 and 4y = –2x+1

  4. #4 peak

    of most nouns.⏎Add -es if the noun ends in -s, -sh, -ch, or -

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. L12:893 excites
  2. L10:217 excites
  3. L6:32 excites
  4. L11:3068 excites
  5. L9:1607 suppresses
  6. L9:30 suppresses
  7. L7:712 excites
  8. L7:21 excites
  9. L12:707 suppresses
  10. L12:92 suppresses
  11. L5:8 excites
  12. L6:561 excites
  13. L9:280 suppresses
  14. L4:2 suppresses
  15. L7:138 excites 'to' after verbs like prove or time
  16. L12:431 excites
  17. L11:300 suppresses
  18. L12:1468 suppresses
  19. L12:1112 excites
  20. L10:105 excites
  21. L7:116 excites
  22. L8:9 weakly excites
  23. L7:19 weakly suppresses
  24. L7:173 weakly excites

Feeds later layers

  1. L15:286 strongly suppresses
  2. L18:93 strongly suppresses
  3. L19:413 strongly excites
  4. L14:1091 strongly excites
  5. L14:348 suppresses
  6. L14:453 excites
  7. L19:223 suppresses
  8. L19:335 suppresses
  9. L15:1601 excites
  10. L17:694 excites
  11. L17:202 excites
  12. L17:146 excites period after abbreviations and dates
  13. L16:323 suppresses
  14. L17:354 suppresses
  15. L20:79 suppresses
  16. L17:375 excites
  17. L16:662 excites
  18. L20:71 suppresses
  19. L16:1590 suppresses
  20. L15:316 suppresses
  21. L19:182 excites
  22. L16:2412 suppresses
  23. L14:143 suppresses 'the' inside journal and institution names
  24. L22:160 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 -1.28 and -1.203 values between -1.06 and -0.997 values between -0.99 and -0.9118 values between -0.91 and -0.8428 values between -0.84 and -0.7758 values between -0.77 and -0.6989 values between -0.69 and -0.62117 values between -0.62 and -0.55138 values between -0.55 and -0.48209 values between -0.48 and -0.40225 values between -0.40 and -0.33270 values between -0.33 and -0.26269 values between -0.26 and -0.19358 values between -0.19 and -0.11399 values between -0.11 and -0.04604 values between -0.04 and 0.03442 values between 0.03 and 0.11302 values between 0.11 and 0.18174 values between 0.18 and 0.25111 values between 0.25 and 0.3238 values between 0.32 and 0.4022 values between 0.40 and 0.478 values between 0.47 and 0.541 values between 0.54 and 0.624 values between 0.62 and 0.691 values between 0.76 and 0.831 values between 0.83 and 0.911 values between 0.91 and 0.982 values between 1.05 and 1.121 values between 1.12 and 1.202 values between 1.20 and 1.273 values between 1.27 and 1.341 values between 1.42 and 1.491 values between 1.56 and 1.63offnegativepositiveoften

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. 的最后一
  2. anned
  3. emo
  4. esian
  5. -lat
  6. 已久的
  7. 的第一
  8. arts
  9. 的认可
  10. edral

Suppresses

  1. "|
  2. ほうが
  3. iscopal
  4. '|
  5. 続きを読む
  6. getParameter
  7. '%(
  8. respect
  9. "%(
  10. indices