Neuron Explorer

Qwen3-0.6B

the word 'also' inside a sentence

A neuron in layer 8 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 lights up on the word "also" when it sits between a subject and its verb or predicate, as in "is also clear" or "are also designed." The negative side fires on "and" or "to" linking two items. Turning it up made the model's continuations wander without a clear shared theme.

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

    smoking, without any negative social or economic consequences.⏎The link between secondhand smoke and disease is well known, and the connection to cardiovascular-related disability and death is also clear. In fact, in a study from Helena, Mont., last year,

  2. #2 peak

    is diverse. For instance, the religion consists of two major branches; Shinto and Buddhism. The country holds many national celebrations in a year. Social classes are also stratified. Gender roles and social activities in Japan follow an order. As an

  3. #3 peak

    such as the 1D finite square potential well. Because the well is symmetrical about its midpoint, it follows that the electron's proabability density is also symmetrical about the midpoint (i.e. even). Recall that the probability density

  4. #4 peak

    burn or be damaged!⏎Precaution 1: When traveling⏎US sockets are generally 110V, therefore electrical appliances plugs for the US are also designed to match the voltage of 110. However, the socket volt

  5. #5 peak

    . Nutrition mistakes.⏎Chervil contains essential volatile oils which determine the flavor of the plant. Wild chervil also contains bitter compounds, glycosides, carotene, vitamin C, and

  6. #6 peak

    summers, when people beat the heat with chilled drinks.⏎AdvertisementFurther sinusitis is always on the rise due to bacteria and fungus in the atmosphere. Traffic pollution is also a big culprit. It is better to seek immediate medical attention when people suffer persistent

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: A square with the diagonals... the point... and the point... and the sum is equal to? This is a technique used to find the sum... the product... and the result. The figure is a... or... the sum...

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 restrictions of a great city. Autumn and winter, spring and summer, come to every window. […] The garden is the medium and the agency

  2. #2 peak

    ), describes the “proper” social relationships and the set of rituals which regulate them. The fundamental social relationships are those between the emperor and his ministers, between father and son, between husband and wife

  3. #3 peak

    either top to bottom or bottom to top). If the trimmer is placed parallel to the nail (cutting from

  4. #4 peak

    may not be exchanged for money, for it is their property for ever.⏎He is not to substitute it or exchange it the good with the bad or the bad with the good. If he ever makes an exchange of an animal for an animal,

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. L7:148 excites
  2. L7:94 suppresses
  3. L7:85 excites
  4. L5:51 suppresses
  5. L7:75 excites
  6. L7:349 excites
  7. L6:1592 excites
  8. L7:222 suppresses
  9. L6:445 excites
  10. L6:743 suppresses
  11. L6:539 excites
  12. L6:265 excites
  13. L3:345 suppresses
  14. L6:193 suppresses
  15. L5:419 weakly excites
  16. L5:606 weakly excites
  17. L7:201 weakly excites
  18. L7:2070 weakly excites
  19. L1:17 weakly suppresses
  20. L5:1510 weakly excites
  21. L6:192 weakly excites
  22. L5:141 weakly excites
  23. L0:814 weakly suppresses
  24. L7:309 weakly excites

Feeds later layers

  1. L9:337 suppresses
  2. L10:2162 suppresses
  3. L10:453 excites
  4. L12:198 excites
  5. L10:133 suppresses
  6. L9:412 excites
  7. L10:1396 weakly suppresses
  8. L13:2067 weakly suppresses
  9. L9:582 weakly suppresses
  10. L11:738 weakly suppresses
  11. L10:1085 weakly excites
  12. L11:1705 weakly suppresses
  13. L9:695 weakly excites
  14. L11:1885 weakly excites
  15. L19:762 weakly excites
  16. L11:1001 weakly excites
  17. L10:312 weakly excites
  18. L18:1579 weakly excites
  19. L9:1170 weakly suppresses
  20. L10:171 weakly suppresses
  21. L11:2122 weakly suppresses
  22. L18:269 weakly suppresses
  23. L17:2171 weakly suppresses
  24. L10:515 weakly 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 -1.36 and -1.251 values between -0.81 and -0.701 values between -0.70 and -0.593 values between -0.59 and -0.4816 values between -0.48 and -0.37136 values between -0.37 and -0.26452 values between -0.26 and -0.151067 values between -0.15 and -0.041309 values between -0.04 and 0.07602 values between 0.07 and 0.18201 values between 0.18 and 0.2963 values between 0.29 and 0.4015 values between 0.40 and 0.5115 values between 0.51 and 0.624 values between 0.62 and 0.735 values between 0.73 and 0.842 values between 0.95 and 1.062 values between 1.06 and 1.171 values between 1.17 and 1.281 values between 1.28 and 1.392 values between 1.39 and 1.502 values between 1.61 and 1.721 values between 1.72 and 1.831 values between 1.83 and 1.941 values between 1.94 and 2.052 values between 2.61 and 2.721 values between 2.72 and 2.831 values between 2.94 and 3.05offnegativepositiveoften

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. =all
  2. 急
  3. 看出
  4. ]|
  5. {*
  6. ิน
  7. 一下子
  8. 一下子就
  9. inis
  10. :⏎⏎⏎

Suppresses

  1. Rape
  2. arian
  3. ivity
  4. orio
  5. vertis
  6. hart
  7. aria
  8. arma
  9. Matter
  10. emas