Neuron Explorer

Qwen3-0.6B

unclear: word pieces in names and foreign text

A neuron in layer 3 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 fires on fragments in the middle of names and words, often in Dutch, German, Latvian or Greek text. There is no clear pattern beyond that, and turning it up or down did not change the writing in an obvious way.

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

    Houdini’s career led him to collaborate with Scientific American on a lengthy exposé of

  2. #2 peak

    ukken berustende bij de Vereenigde Doopsgezinde Gemeente to Amsterdam, 2 vols. Amsterdam: Uitgegeven en ten geschenke aangeboden door den Kerkeraad d

  3. #3 peak

    Um dies (i.e. Varro's description of the extent of Roma quadrata) mit dem unten erörterten Pomerium und der Auffassung des Dionys. I 88: περιγρ

  4. #4 peak

    products. In the spring of 1934, the Franks r…⏎E-pasta adrese, uz kuru nosūtīt darba saiti:⏎Saite uz darbu:

  5. #5 peak

    ἔλεγεν αὐτοῖς,⏎τὸ σάββατο

  6. #6 peak

    This is the most obvious similarity between the two religions. In fact a Muslim is defined as anyone who says "Th…⏎E-pasta adrese, uz kuru nosūtī

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 thinking about this. I have been thinking about this. I have been thinking about this. I have been thinking about this. I have been thinking about this. I have been thinking about this. I have been thinking

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

    bill to establish a federal holiday in 1894. President Grover Cleveland signed the bill soon afterward, designating the first Monday in September as Labor Day.⏎Learners are Leaders, Inc.’s mission includes uplifting future leaders through

  2. #2 peak

    trade unions was given further impetus in the freedom struggle against British colonialism. One of the earliest trade unions, Tata Workers Union (also alternatively known as Jamshedpur’s Workers’ Union) (http://www.tataworkersunion.org/)

  3. #3 peak

    ⏎Principal researcher: Reyhaneh Bastani⏎Supervisor: Dr. Beaumie Kim⏎P

  4. #4 peak

    of 1857. He is known for his efforts to modernize his country and for restoring Mexico to a Republic.⏎May 1st: Labor Day, commemorates the union movement of the Mexican worker. Article 123

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. L0:1301 suppresses
  2. L0:178 suppresses
  3. L0:1224 suppresses
  4. L0:386 weakly suppresses
  5. L1:651 weakly excites
  6. L2:1334 weakly suppresses
  7. L1:1099 weakly excites
  8. L1:2474 weakly excites
  9. L2:36 weakly excites
  10. L1:2013 weakly excites
  11. L0:894 weakly excites
  12. L1:515 weakly excites 'about' in 'it's not about' phrasing
  13. L1:1795 weakly excites
  14. L1:92 weakly excites
  15. L0:874 weakly excites
  16. L0:2779 weakly suppresses
  17. L2:1899 weakly excites
  18. L1:1665 weakly suppresses
  19. L1:1496 weakly excites
  20. L1:2620 weakly excites
  21. L1:491 weakly excites
  22. L2:645 weakly suppresses
  23. L1:2166 weakly suppresses
  24. L2:1882 weakly suppresses

Feeds later layers

  1. L4:528 weakly excites
  2. L6:322 weakly excites
  3. L12:538 weakly excites
  4. L6:31 weakly excites
  5. L5:146 weakly suppresses
  6. L24:190 weakly excites
  7. L27:351 weakly suppresses
  8. L16:2481 weakly excites
  9. L25:587 weakly suppresses
  10. L7:2560 weakly suppresses
  11. L6:574 weakly suppresses
  12. L15:2423 weakly excites
  13. L6:1667 weakly suppresses
  14. L24:1218 weakly suppresses
  15. L7:3 weakly excites
  16. L19:493 weakly excites
  17. L8:997 weakly suppresses
  18. L8:2236 weakly suppresses
  19. L10:214 weakly suppresses
  20. L5:1799 weakly excites
  21. L14:1909 weakly suppresses
  22. L14:2801 weakly suppresses
  23. L10:1565 weakly excites
  24. L23:788 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 -0.56 and -0.531 values between -0.53 and -0.512 values between -0.46 and -0.432 values between -0.43 and -0.402 values between -0.40 and -0.384 values between -0.38 and -0.351 values between -0.35 and -0.3311 values between -0.33 and -0.3010 values between -0.30 and -0.2720 values between -0.27 and -0.2525 values between -0.25 and -0.2240 values between -0.22 and -0.2045 values between -0.20 and -0.17109 values between -0.17 and -0.14134 values between -0.14 and -0.12175 values between -0.12 and -0.09254 values between -0.09 and -0.06361 values between -0.06 and -0.04482 values between -0.04 and -0.01613 values between -0.01 and 0.01473 values between 0.01 and 0.04362 values between 0.04 and 0.07257 values between 0.07 and 0.09172 values between 0.09 and 0.12117 values between 0.12 and 0.1577 values between 0.15 and 0.1749 values between 0.17 and 0.2046 values between 0.20 and 0.2216 values between 0.22 and 0.2516 values between 0.25 and 0.2810 values between 0.28 and 0.308 values between 0.30 and 0.335 values between 0.33 and 0.352 values between 0.35 and 0.383 values between 0.38 and 0.411 values between 0.41 and 0.431 values between 0.43 and 0.461 values between 0.46 and 0.48offnegativepositiveoften

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. 他们会
  2. 若您
  3. ento
  4. UPER
  5. getResources
  6. 猖
  7. onward
  8. его
  9. fasc
  10. Fasc

Suppresses

  1. ocha
  2. Necklace
  3. ấm
  4. ellig
  5. proble
  6. 晕
  7. unya
  8. 糖
  9. 饽
  10. blem