Neuron Explorer

Qwen3-0.6B

unclear: pieces inside numbers and blanks

A neuron in layer 9 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron is weak and hard to read. It fires on odd spots such as a single digit inside a number, a blank line, or a comma in a big number. Nothing about it is clearly about one topic, and turning it up did not bring out a clear theme.

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

    128.0/17 SUBNET3 : 10101100.00010001.0yyyyyyy.yyyyyyyy => 172.17.0.0

  2. #2 peak

    were 1, 2 and 134. All these vehicles were fitted with ballast ploughs.⏎N 1

  3. #3 peak

    1, 1/4 turn clockwise and then back 2_______.⏎5.1/4 turn clockwise, forward 3, 3

  4. #4 peak

    and “motto” theme, but does not stay there long. The tumultuous Allegro begins in D minor, a key as far as possible from A-flat (Elgar wrote to his friend August Jaeger that he had “thrown over

  5. #5 peak

    48x2048 pixels,⏎be sure to set camera to this or lower⏎- Handheld GPS,⏎preferably one with 100’ or better accuracy that allows⏎marking locations and naming them⏎Accuracy is important

  6. #6 peak

    ||Water (%)||1.57% (1,014 km2)|⏎|-||2011 estimate||2,22

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 opposite of the same In the opposite of the same In the opposite of the same In the opposite of the same In the opposite of the same In the opposite of the same In the opposite of the same

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

    plumage with a black back.⏎- ‘Neophytes at duck identification often confuse a bufflehead with a male hooded merganser when the latter's crest is raised showing a large patch of white feathers.’⏎- ‘These

  2. #2 peak

    a familiar spirit.⏎How do we reconcile these apparently conflicting statements? We do so by recognizing that the latter statement is the figure of speech “metonymy of the subject.” The actual

  3. #3 peak

    (1) military operations, (2) weapons technology, (3) diplomatic activities, (4) intelligence activities, and (5) cryptology. The latter two areas might be considered to be special parts of the first three areas. That

  4. #4 peak

    , Guatemala, Florida, and France between 1789 and 1809.2 Much remains to be known about the St. [End Page 37] Dominguan diaspora in France and in the circum-Car

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. L6:267 strongly excites
  2. L4:77 suppresses
  3. L4:325 excites
  4. L5:51 excites
  5. L8:1671 suppresses
  6. L2:51 suppresses
  7. L6:763 excites
  8. L7:50 suppresses
  9. L5:851 excites
  10. L6:355 suppresses
  11. L4:436 suppresses
  12. L5:200 suppresses
  13. L1:46 excites
  14. L8:47 suppresses
  15. L6:955 excites
  16. L6:180 suppresses
  17. L8:57 excites
  18. L5:1510 suppresses
  19. L1:107 excites
  20. L7:475 excites
  21. L7:1866 excites
  22. L3:16 excites
  23. L1:1371 suppresses
  24. L0:811 suppresses opening parenthesis after a bracket or name

Feeds later layers

  1. L12:1209 excites
  2. L15:483 excites
  3. L12:198 excites
  4. L11:530 suppresses
  5. L10:2162 suppresses
  6. L12:1483 excites
  7. L13:932 suppresses second digit of a year range
  8. L10:1024 suppresses
  9. L13:452 excites
  10. L13:1020 excites
  11. L10:1396 suppresses
  12. L12:532 suppresses
  13. L14:477 excites
  14. L12:2301 suppresses
  15. L12:2783 suppresses
  16. L12:2042 suppresses
  17. L11:1705 suppresses
  18. L10:1062 excites
  19. L10:506 suppresses
  20. L13:1783 suppresses
  21. L10:171 suppresses
  22. L13:1497 excites
  23. L14:566 suppresses
  24. L14:845 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 -0.92 and -0.872 values between -0.87 and -0.821 values between -0.82 and -0.772 values between -0.77 and -0.724 values between -0.72 and -0.6714 values between -0.67 and -0.6210 values between -0.62 and -0.5716 values between -0.57 and -0.5224 values between -0.52 and -0.4732 values between -0.47 and -0.4241 values between -0.42 and -0.3762 values between -0.37 and -0.3274 values between -0.32 and -0.27104 values between -0.27 and -0.22143 values between -0.22 and -0.17230 values between -0.17 and -0.12331 values between -0.12 and -0.07596 values between -0.07 and -0.021227 values between -0.02 and 0.03449 values between 0.03 and 0.08194 values between 0.08 and 0.13141 values between 0.13 and 0.1879 values between 0.18 and 0.2336 values between 0.23 and 0.2828 values between 0.28 and 0.3317 values between 0.33 and 0.3810 values between 0.38 and 0.4313 values between 0.43 and 0.489 values between 0.48 and 0.538 values between 0.53 and 0.581 values between 0.58 and 0.634 values between 0.63 and 0.682 values between 0.73 and 0.782 values between 0.78 and 0.831 values between 1.03 and 1.08offnegativepositiveoften

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. eled
  2. eler
  3. 岁月
  4. 三项
  5. очных
  6. очный
  7. urable
  8. вел
  9. qué
  10. /animate

Suppresses

  1. 更深
  2. 他还
  3. đáo
  4. adipiscing
  5. 他又
  6. оборот
  7. brighter
  8. höher
  9. Tat
  10. arat