Neuron Explorer

Qwen3-0.6B

unclear: commas after dates and clauses

A neuron in layer 19 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 is weak and hard to pin down. Its strongest moments come at commas after a date or a short phrase in news-style text, such as "Thu, 3 Jun". Turning it up or down did not produce a clear change in the 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

    17:55 EST⏎- ASU instrument on NASA rover helps identify outcrop of long-sought rare rock on MarsThu, 3 Jun 2010, 14:52:

  2. #2 peak

    Comments are closed.<|endoftext|>The bittersweet truth about how taste works⏎Scientists have discovered an ion channel that allows taste buds to communicate with the brain.⏎Wed, Mar 06, 2013 at 03:2

  3. #3 peak

    program via a distance learning course for the first time. You can even watch a promotional video from a North Korean web site. (It helps if you speak Korean, which I don

  4. #4 peak

    tumor dies. In the period of a few weeks to months, it forms back. In rare cases, the tumor swells after treatment. This is mostly temporary and is easy to control.⏎How often is a cyberknife treatment repeatable?⏎As

  5. #5 peak

    Philadelphia.⏎After graduating, she practiced in Germany where she also supported women doctors, helping to establish a women’s clinic and hospital.<|endoftext|>It's Hiking Season, But How Many Calories Are You Really Burning?⏎When the weather gets colder, one

  6. #6 peak

    that you are ready to share it. First a few important points:⏎Now you have the modification as a source file. Sending the source file will not be helpful, as probably someone else has done other modifications to the original file in the meantime.

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: If, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, 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

    keeps the form of Amor fati (Love of Fate) through the important element of Nietzsche's philosophy, the "will to power" (der Wille zur Macht), the basis of human behavior, influenced by the Will to Live of

  2. #2 peak

    , 168 note) who remarks: — "Um dies (i.e. Varro's description of the extent of Roma quadrata) mit dem unten erörterten Pomerium und der Auffassung des Dion

  3. #3 peak

    zeichnung der Linie des Pomeriums, zweitens der parallelen Linie der Befestigung der Arx."⏎

  4. #4 peak

    ing Spiritualism and Materialism’, Ludwig Andreas Feuerbach wrote- “Der Mensch ist, was er ißt.” (Man is what he eats) and by

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. L17:6 strongly suppresses
  2. L18:1651 strongly suppresses
  3. L18:823 strongly excites
  4. L17:14 suppresses
  5. L18:247 excites
  6. L17:18 excites unclear: mixed mid-sentence words
  7. L18:1956 suppresses
  8. L18:1444 excites
  9. L17:714 excites
  10. L18:560 excites
  11. L18:624 excites
  12. L18:765 suppresses
  13. L5:86 suppresses
  14. L16:2749 weakly excites
  15. L18:1018 weakly excites
  16. L6:7 weakly suppresses
  17. L18:569 weakly excites
  18. L18:477 weakly suppresses
  19. L1:1049 weakly excites
  20. L18:320 weakly excites
  21. L12:15 weakly excites
  22. L2:126 weakly excites
  23. L5:28 weakly excites
  24. L3:464 weakly excites

Feeds later layers

  1. L20:661 strongly suppresses
  2. L20:343 strongly suppresses
  3. L20:351 strongly excites
  4. L24:43 strongly suppresses
  5. L23:253 strongly suppresses
  6. L25:198 suppresses
  7. L24:512 suppresses 'the' inside organisation and convention names
  8. L21:185 excites
  9. L20:1198 suppresses
  10. L23:278 suppresses
  11. L25:594 suppresses
  12. L25:67 suppresses
  13. L25:406 suppresses
  14. L27:535 suppresses
  15. L27:133 suppresses
  16. L27:12 suppresses
  17. L27:16 suppresses
  18. L26:277 suppresses
  19. L26:161 suppresses
  20. L27:90 suppresses
  21. L23:105 excites
  22. L20:940 suppresses
  23. L27:293 suppresses
  24. L27:406 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 -8.02 and -7.771 values between -1.08 and -0.847 values between -0.59 and -0.3470 values between -0.34 and -0.091388 values between -0.09 and 0.161384 values between 0.16 and 0.40770 values between 0.40 and 0.65233 values between 0.65 and 0.9040 values between 0.90 and 1.1511 values between 1.15 and 1.392 values between 1.39 and 1.641 values between 1.64 and 1.89offnegativepositiveoften

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. entic
  2. base
  3. borough
  4. back
  5. cap
  6. logue
  7. ina
  8. van
  9. mana
  10. rea

Suppresses

  1. Releases
  2. Vertical
  3. *out
  4. Experiment
  5. ernen
  6. 不解
  7. ёр
  8. ,ev
  9. 陌生人
  10. ез