Neuron Explorer

Qwen3-0.6B

passive 'formed by' and 'based on' phrasing

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 fires on the small word right after a noun that starts a 'made by' or 'based on' phrase, like 'formed by', 'generated', or 'came from'. Turning it up did not produce a clear theme in the model's writing.

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

    additional requirements that will protect bees. EPA will consider the other uses at a later date as data become available to support those uses. EPA made this decision after careful consideration of public comments and supporting science.⏎E

  2. #2 peak

    pouring of similarly signed romantic notes in subsequent centuries.⏎Nineteenth Century<|endoftext|>Icelandic Sheep Breeders of North America⏎The following rules were formed by the late Dr. Halldor Palsson who was the leading sheep specialist in

  3. #3 peak

    Earth but still might contain life.<|endoftext|>New guidelines have been launched for educators, using of electronic communication like social media.⏎The guidelines were created by a working group formed by the

  4. #4 peak

    activity is designed to have students analyze energy use, cost, and source patterns from household to regional scales and relate these patterns to CO2 emissions. This idea was generated at ...⏎Learn more about this review process.⏎Mentos and soda eruptions

  5. #5 peak

    new area of dark patches that covers some 34,000 square kilometres.⏎The new dark patches are most likely lakes of hydrocarbons, created by rain that poured from summer clouds, says Turtle. "It's the first time

  6. #6 peak

    civil war with the Nationalists under Chiang Kai-shek, in which millions died.⏎Chiang and the remains of his government fled to Taiwan that December after a last stand in

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: what is the point of being a good and useful person in the future? So, considering the future, what can be done or not done in a professional, personal, or societal context to maintain or maintain a good standing? So, 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

    various shades of colours. These LEDs have four legs. … To control colours on an RGB LED we would use pulse width modulation, or PWM for short.⏎Do LEDs use PWM?⏎LEDs rely on PWM for maintaining their voltage.⏎

  2. #2 peak

    months. In fact, feeling blue during the late fall and throughout the winter is so common that it has a name: seasonal affective disorder. SAD for short, this condition is often called the Winter Blues. But if you think these feelings

  3. #3 peak

    well-known way to initiate meanings out-of acronyms, which happen to be abbreviations noticable due to the fact words: RAID, and that is short for redundant selection of separate (otherwise cheaper) discs, identifies a pc shops program that

  4. #4 peak

    the primary problem lies with the heart muscle cells themselves; these diseases are called cardiomyopathies. Different cardiomyopathies are caused by different cellular defects, but most

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. L7:152 suppresses
  2. L5:2925 excites
  3. L8:753 weakly suppresses
  4. L8:1195 weakly suppresses
  5. L6:733 weakly suppresses
  6. L6:1949 weakly excites
  7. L3:2806 weakly suppresses
  8. L7:2726 weakly suppresses
  9. L1:2388 weakly excites
  10. L6:1654 weakly excites
  11. L6:1254 weakly excites
  12. L6:1975 weakly excites
  13. L7:773 weakly suppresses
  14. L6:2089 weakly suppresses
  15. L6:897 weakly excites
  16. L5:83 weakly suppresses
  17. L7:1880 weakly suppresses
  18. L7:633 weakly suppresses
  19. L1:2141 weakly excites
  20. L2:1653 weakly excites
  21. L2:2416 weakly excites
  22. L8:1735 weakly suppresses the word 'National' in park names
  23. L7:1326 weakly suppresses
  24. L7:912 weakly excites

Feeds later layers

  1. L10:2232 weakly excites
  2. L13:236 weakly excites
  3. L12:1405 weakly excites
  4. L11:1367 weakly excites
  5. L15:1642 weakly excites
  6. L25:236 weakly excites
  7. L26:1024 weakly suppresses
  8. L11:896 weakly suppresses comma closing an introductory phrase
  9. L15:1519 weakly excites
  10. L10:396 weakly excites
  11. L17:24 weakly excites
  12. L11:992 weakly excites
  13. L17:2634 weakly suppresses
  14. L17:1761 weakly suppresses
  15. L11:1446 weakly excites
  16. L16:505 weakly excites
  17. L25:2054 weakly suppresses
  18. L21:2553 weakly excites
  19. L21:2755 weakly suppresses
  20. L17:2262 weakly suppresses
  21. L15:2757 weakly suppresses
  22. L11:741 weakly suppresses
  23. L12:29 weakly excites
  24. L10:1375 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.

4 values between -0.66 and -0.593 values between -0.59 and -0.527 values between -0.52 and -0.456 values between -0.45 and -0.3841 values between -0.38 and -0.3194 values between -0.31 and -0.24238 values between -0.24 and -0.16428 values between -0.16 and -0.09778 values between -0.09 and -0.021080 values between -0.02 and 0.05548 values between 0.05 and 0.12328 values between 0.12 and 0.19196 values between 0.19 and 0.2675 values between 0.26 and 0.3333 values between 0.33 and 0.4015 values between 0.40 and 0.4714 values between 0.47 and 0.547 values between 0.54 and 0.611 values between 0.61 and 0.681 values between 0.68 and 0.751 values between 0.75 and 0.821 values between 0.89 and 0.971 values between 0.97 and 1.042 values between 1.04 and 1.114 values between 1.25 and 1.321 values between 1.46 and 1.531 values between 2.09 and 2.16offnegativepositiveoften

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. stadt
  2. ient
  3. 幕后
  4. pon
  5. arios
  6. stant
  7. 另有
  8. hd
  9. consulted
  10. due

Suppresses

  1. /problem
  2. tighter
  3. ck
  4. ложений
  5. cker
  6. ück
  7. ivicrm
  8. principalColumn
  9. (vertical
  10. isque