Neuron Explorer

Qwen3-0.6B

'and' or 'or' in a list of three or more

A neuron in layer 14 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 connecting word 'and' or 'or' or a comma near the end of a list of several items, such as 'endangered, vulnerable, and threatened'. Its negative side is weak and mixed. Turning it up made the model's writing repetitive and mathematical, which does not clearly match the trigger.

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

    Bush Administration. This tradition extends back to America’s original sin: the ethnic cleansing of America’s indigenous people.⏎To take just one example haters, killers, and duelers, such as Andrew Jackson, president of the United States when the Cherokee

  2. #2 peak

    materials into landfills that may be used on other projects or donated to Habitat Restore.⏎- Transportation jobs - If you work for a trucking company, a city or state transportation department or any other organization that uses

  3. #3 peak

    , and stem are not toxic to felines, making the vegetable a great supplement to their diet. Celery chipmunks and squirrels endangered, vulnerable, and threatened species flower gardening african wild dogs animal life zoos cats (felines)

  4. #4 peak

    mode of existence of a society. These⏎change as all societies are subject to⏎change. There is no permanent society or culture.⏎The cultural balance, pattern or synthesis that exists in a society at a given historical stage is nothing but the unity

  5. #5 peak

    process is more formally known as tearing apart. Place your book on a flat surface, and with a sharp knife, cut right through the endsheet, mull and the covering material as depicted in figure 180. While doing this,

  6. #6 peak

    ’s natural release of melatonin. For instance, a light from a charger – keep all electronics out of your bedroom!⏎- Use your bed for sleep, reading and sex only (not for

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: I have a, the, and, the, and, the, and, the, and, the, and, the, and, the, and, the, and, the, and, the, and, the, and,

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

    algebra problems, finding the roots of nonlinear equations, functions written in Octave's own language, or by using dynamically loaded modules written in C, C++, Fortran, or other languages.⏎Donations to support the software can be made at https

  2. #2 peak

    ⏎The main crack<|endoftext|>analysis of algoritms⏎robert.kern at gmail.com⏎Fri Sep 10 00:10:22 CEST 2010⏎On 9/9/10

  3. #3 peak

    A New, Authentic and Complete Collection of Voyages Round the World…Complete Historical Account of Captain Cook's First, Second, Third and Last Voyages… Edited by George William Anderson and published by Alexander Hogg

  4. #4 peak

    world?⏎- How can I complain to police in Karnataka online?⏎- What does an eagle with two heads mean?⏎- Did the Hittites invade Egypt?⏎- What does the word Hittite mean?⏎- Who were the Hittite

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. L11:1595 strongly excites
  2. L9:102 excites
  3. L7:307 suppresses
  4. L12:1243 excites
  5. L7:284 suppresses
  6. L13:1554 excites
  7. L13:372 suppresses
  8. L6:906 excites
  9. L13:1627 suppresses
  10. L13:1020 suppresses
  11. L8:100 excites
  12. L13:1469 excites
  13. L9:275 excites
  14. L13:926 excites
  15. L8:558 excites
  16. L11:228 excites
  17. L12:501 suppresses
  18. L13:750 suppresses
  19. L12:1385 excites
  20. L7:675 excites
  21. L13:876 excites
  22. L4:1544 suppresses
  23. L6:2947 weakly suppresses
  24. L10:1948 weakly excites

Feeds later layers

  1. L16:660 strongly suppresses
  2. L17:2155 excites
  3. L16:744 excites
  4. L15:1708 suppresses
  5. L16:509 excites
  6. L15:361 suppresses
  7. L15:1213 suppresses
  8. L17:2413 suppresses
  9. L17:1880 weakly excites
  10. L15:2954 weakly excites
  11. L19:2416 weakly excites
  12. L16:2160 weakly excites
  13. L17:2332 weakly excites 'and' joining two clauses
  14. L17:2084 weakly excites
  15. L20:1024 weakly suppresses
  16. L16:767 weakly suppresses
  17. L16:474 weakly suppresses
  18. L16:2417 weakly excites
  19. L17:274 weakly excites
  20. L27:340 weakly suppresses parenthesis introducing an acronym
  21. L18:2342 weakly suppresses
  22. L22:373 weakly excites
  23. L17:1567 weakly suppresses
  24. L17:1300 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.

14 values between -1.46 and -1.02120 values between -1.02 and -0.591139 values between -0.59 and -0.152506 values between -0.15 and 0.2865 values between 0.28 and 0.7217 values between 0.72 and 1.155 values between 1.15 and 1.597 values between 1.59 and 2.025 values between 2.02 and 2.462 values between 2.46 and 2.895 values between 2.89 and 3.331 values between 3.33 and 3.761 values between 3.76 and 4.203 values between 4.20 and 4.631 values between 4.63 and 5.072 values between 5.07 and 5.503 values between 5.94 and 6.373 values between 6.81 and 7.242 values between 7.24 and 7.681 values between 7.68 and 8.111 values between 8.11 and 8.551 values between 8.55 and 8.981 values between 9.85 and 10.292 values between 10.29 and 10.721 values between 15.50 and 15.94offnegativepositiveoften

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. 这个游戏
  2. 这款游戏
  3. "](
  4. [axis
  5. .parts
  6. avig
  7. ]byte
  8. distributed
  9. 这个名字
  10. released

Suppresses

  1. ório
  2. O
  3. ritos
  4. allas
  5. izons
  6. esh
  7. ueil
  8. orio
  9. ning
  10. yssey