Neuron Explorer

Qwen3-0.6B

the letter 'l' inside words

A neuron in layer 26 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 letter l when it appears in the middle or start of a word, such as in lazaretto, Lorikeet or Llangynhafal. Turned up, the model produced repetitive, garbled text, so the steering output does not show a clear theme.

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

    of St Basil originally selected a name for their leper hospital. The name eventually gave rise to the terms "lazar", meaning a leper, and "lazaretto", a leprosarium, or Leper House. The

  2. #2 peak

    ⏎Lorikeet feeding habits⏎Lorikeets are brightly coloured parrots of the family Psittacidae and sub-family Loriinae (l

  3. #3 peak

    had gradually swell In his mind. The chronological accounts testify his previous encounters from the very beginning of the narrative “My first mistress was a woman of… To “l had resided but a short time in Baltimore” to “Very soon after I

  4. #4 peak

    a restaurant for leftovers or as a takeout container from a fast food place⏎- egg cartons – either the styrofoam ones or cardboard ones (latter is preferable as you can plant these in the ground and they will rot away

  5. #5 peak

    , having outlived the other apostles — all of whom, according to legend/history<|endoftext|>The

  6. #6 peak

    ancestors. This work of art belongs to the old time of Koh Kok Thlork.⏎After the year 1979, the authority of the Take

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: bothogogogogogogogogogogogogogogogogogogogogogogogogogogogogogogogogogogogogogogogogogogogogogogog

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

    Equine Granulocytic Anaplasmosis (Ehrlichiosis)⏎|Deer (or Black-Legged) Tick⏎||Tear drop-shaped; the size of a pop

  2. #2 peak

    . With time we check more numbers and become more convinced that the conjecture is true. Currently, it has been checked up to the power 4.6 ⋅<|endoftext|>Picture Yourself in Orbit: Are You Really Weightless?⏎An object

  3. #3 peak

    0.97. The expected number of powers of 2 without zeroes starting from the power N is cxN/(1-x) ≈ 40 ⋅ 0.97N.⏎Let us look at A00

  4. #4 peak

    secretpassword';⏎mysql> GRANT ALL PRIVILEGES ON database_name.* TO user_name@localhost;⏎mysql> FLUSH PRIVILEGES;⏎Step 5: Installation and Configuration of Drupal⏎Open

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. L19:322 strongly excites
  2. L17:134 strongly suppresses
  3. L21:325 excites
  4. L18:420 excites
  5. L20:1710 suppresses
  6. L16:117 suppresses
  7. L0:94 excites
  8. L2:1518 suppresses
  9. L19:1639 excites
  10. L23:1760 suppresses
  11. L22:3039 excites
  12. L6:2953 excites
  13. L24:2164 excites
  14. L5:1294 excites
  15. L16:55 suppresses
  16. L25:1316 suppresses
  17. L0:662 suppresses
  18. L2:1866 excites
  19. L1:23 excites
  20. L23:2074 suppresses
  21. L16:177 excites zero digit inside dates and codes
  22. L10:1349 weakly excites
  23. L7:904 weakly excites
  24. L1:31 weakly suppresses

Feeds later layers

  1. L27:1827 weakly suppresses
  2. L27:1602 weakly suppresses
  3. L27:2749 weakly suppresses
  4. L27:2772 weakly suppresses
  5. L27:2906 weakly suppresses the letter r in the middle of words
  6. L27:1897 weakly excites
  7. L27:914 weakly excites
  8. L27:2899 weakly suppresses
  9. L27:1528 weakly suppresses
  10. L27:2550 weakly suppresses
  11. L27:698 weakly suppresses
  12. L27:1084 weakly excites
  13. L27:2962 weakly suppresses
  14. L27:2421 weakly excites
  15. L27:556 weakly excites
  16. L27:1164 weakly excites
  17. L27:2777 weakly suppresses
  18. L27:1604 weakly suppresses
  19. L27:2337 weakly excites
  20. L27:418 weakly excites
  21. L27:1250 weakly suppresses
  22. L27:2618 weakly excites
  23. L27:1265 weakly excites
  24. L27:3011 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 -23.55 and -22.261 values between -13.25 and -11.961 values between -6.81 and -5.521 values between -5.52 and -4.233 values between -4.23 and -2.956 values between -2.95 and -1.66103 values between -1.66 and -0.373772 values between -0.37 and 0.929 values between 0.92 and 2.203 values between 2.20 and 3.491 values between 3.49 and 4.781 values between 4.78 and 6.071 values between 7.35 and 8.641 values between 8.64 and 9.931 values between 9.93 and 11.221 values between 15.08 and 16.371 values between 22.80 and 24.091 values between 26.67 and 27.95offnegativepositiveoften

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. al
  2. alink
  3. 恰恰
  4. 至於
  5. enting
  6. 每人
  7. 用微信
  8. quello
  9. 煦
  10. indemn

Suppresses

  1. og
  2. ine
  3. oad
  4. ess
  5. odash
  6. ike
  7. iations
  8. ий
  9. ogh
  10. igkeit