Neuron Explorer

Qwen3-0.6B

unclear: period at fragment start

A neuron in layer 1 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 a full stop, but the examples are all at the very start of a text fragment, so it may be reacting to where the text was cut rather than to any meaning. The pattern is not clear, and the steering outputs do not explain it.

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

    . For example, it’s a good idea to place the loading and unloading bay

  2. #2 peak

    . of the Afro-Asiatic language family that includes Hebrew, Arabic, Ar

  3. #3 peak

    . The hotel also had several amenities, including a casino, a golf course, and

  4. #4 peak

    . If there are no local brands to satisfy a need, then consumers will purchase foreign

  5. #5 peak

    . Election day was largely peaceful and approximately 81% of voters cast ballots,

  6. #6 peak

    . The new method, published today in the journal Nature, makes it possible to store

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 been reading about some books, and I have been thinking about my life. I am reading a book called "the journey" and I have been reading about the life and the life. I have been reading about my life. I

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

    OBJECT.⏎SUBTRACTIVE SPACES AND THE DEEP WALL.⏎INDEX OF BUILDINGS

  2. #2 peak

    -blown AIDS and death from AIDS (Goedert et<|endoftext|>IJC Releases

  3. #3 peak

    Описание презентации Презентация london-history-12162

  4. #4 peak

    -smallest county in Utah by land area and third-smallest by total area.⏎The

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. L0:1224 suppresses
  2. L0:178 suppresses
  3. L0:1301 suppresses
  4. L0:2877 suppresses
  5. L0:552 excites
  6. L0:1780 weakly suppresses
  7. L0:2857 weakly suppresses
  8. L0:2242 weakly excites
  9. L0:2785 weakly suppresses
  10. L0:1945 weakly suppresses
  11. L0:2844 weakly excites
  12. L0:2733 weakly excites
  13. L0:2120 weakly suppresses
  14. L0:1 weakly suppresses
  15. L0:1118 weakly suppresses
  16. L0:2440 weakly suppresses
  17. L0:1793 weakly excites
  18. L0:2583 weakly suppresses
  19. L0:40 weakly excites 'of' in 'sort of' hedges
  20. L0:2589 weakly excites
  21. L0:2998 weakly suppresses
  22. L0:2416 weakly excites
  23. L0:2334 weakly excites
  24. L0:1378 weakly excites

Feeds later layers

  1. L20:70 suppresses
  2. L24:2780 excites
  3. L2:77 weakly excites
  4. L21:298 weakly excites
  5. L22:813 weakly excites
  6. L15:727 weakly suppresses
  7. L27:664 weakly excites
  8. L19:2100 weakly suppresses
  9. L9:250 weakly excites
  10. L22:177 weakly suppresses
  11. L26:2830 weakly excites
  12. L27:224 weakly excites
  13. L27:411 weakly excites
  14. L27:169 weakly excites
  15. L16:791 weakly suppresses
  16. L19:224 weakly suppresses
  17. L12:54 weakly suppresses
  18. L23:335 weakly excites
  19. L5:2120 weakly excites
  20. L16:1114 weakly suppresses
  21. L27:228 weakly excites
  22. L27:293 weakly excites
  23. L26:1098 weakly excites
  24. L19:1399 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.

1 values between -0.41 and -0.371 values between -0.23 and -0.195 values between -0.19 and -0.168 values between -0.16 and -0.1218 values between -0.12 and -0.0962 values between -0.09 and -0.05291 values between -0.05 and -0.022291 values between -0.02 and 0.02913 values between 0.02 and 0.05194 values between 0.05 and 0.0959 values between 0.09 and 0.1218 values between 0.12 and 0.169 values between 0.16 and 0.1911 values between 0.19 and 0.239 values between 0.23 and 0.264 values between 0.30 and 0.334 values between 0.33 and 0.374 values between 0.37 and 0.413 values between 0.41 and 0.441 values between 0.51 and 0.551 values between 0.93 and 0.971 values between 0.97 and 1.00offnegativepositiveoften

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. owi
  2. ican
  3. iances
  4. icted
  5. deceased
  6. sterol
  7. entlich
  8. letal
  9. weed
  10. iane

Suppresses

  1. 给孩子
  2. pliers
  3. akah
  4. 本质
  5. ono
  6. 放大
  7. yes
  8. 金刚
  9. nhóm
  10. 只知道