Neuron Explorer

Qwen3-0.6B

opening parenthesis in lists and asides

A neuron in layer 3 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 responds to an opening bracket '(' in the middle of a sentence, especially in lists or side notes like '(b)' or '(near-field)'. It doesn't seem to respond to the bracket itself so much as to the spot where a list item or aside begins. Turning it up gave text that did not show this pattern.

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

    antibodies and (b) the observation that human CD18 can form chimeric heterod

  2. #2 peak

    protected by thin double-walled carbon nanotubes. The synthesis of very long arrangements is confirmed by a combination of transmission electron microscopy, X-ray diffraction and (near-field) resonance Raman spectroscopy. Our results establish a route for the

  3. #3 peak

    buried for many Ice Age cycles, (but) when the ice advanced over this area, it was essentially frozen to the

  4. #4 peak

    electricity in schools, it is recognised that energy for other purposes such as cooking, hot water, and (in some regions) space heating, also plays an important role in improving educational quality

  5. #5 peak

    0 times faster than any other present method with comparable resolution, (such as super-resolution visible light microscopy or electron microscopy). Professor Hwu estimates that

  6. #6 peak

    digits of 2n. The number of digits can be estimated as n log102. Thus, the probability is about cxn, where c = (10/9)2 ≈ 1.2 and x = (

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 will not. I will not. I will not. I will not. I will not. I will not. I will not. I will not. I will not. I will not. I will not. I will not.

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

    Chegg Com⏎Derivation Of Kir<|endoftext|>The poisonous Gartraps are the bane of the desert and semi desert. They are the same colour as the sand in

  2. #2 peak

    |• Summer (DST)||UTC-5 (CDT)|⏎|GNIS feature ID||685916

  3. #3 peak

    your alcohol and caffeine consumption. Avoid excess caffeine, and no caffeine after 1pm. Aim to consume no more than 1(for females) or 2 (for males) alcoholic beverages per day. Give yourself a break from alcohol at least

  4. #4 peak

    also be empty (Brooks, 2003); complex issues are broken down into broad

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. L0:1252 suppresses
  2. L2:353 weakly excites
  3. L2:2712 weakly excites
  4. L0:1798 weakly suppresses
  5. L1:448 weakly suppresses
  6. L1:1976 weakly suppresses
  7. L0:734 weakly suppresses
  8. L1:23 weakly excites
  9. L1:521 weakly suppresses "hand" in "on the other hand"
  10. L0:11 weakly excites
  11. L0:86 weakly excites
  12. L2:1784 weakly excites
  13. L0:1667 weakly excites
  14. L2:41 weakly suppresses
  15. L1:165 weakly excites
  16. L2:558 weakly excites
  17. L2:1126 weakly suppresses
  18. L0:2329 weakly excites
  19. L0:404 weakly excites
  20. L1:957 weakly suppresses
  21. L2:325 weakly suppresses
  22. L2:2828 weakly excites
  23. L1:499 weakly excites 'in' after success or help verbs
  24. L1:340 weakly excites

Feeds later layers

  1. L6:144 excites
  2. L6:468 weakly suppresses
  3. L6:2957 weakly suppresses
  4. L18:980 weakly excites opening parenthesis after a noun phrase
  5. L7:2907 weakly excites
  6. L6:243 weakly suppresses
  7. L4:1116 weakly excites
  8. L21:2576 weakly suppresses
  9. L24:1809 weakly suppresses
  10. L23:2916 weakly suppresses
  11. L5:2556 weakly excites
  12. L24:505 weakly excites
  13. L24:2157 weakly suppresses
  14. L4:797 weakly excites
  15. L23:2191 weakly excites
  16. L23:2132 weakly suppresses
  17. L11:50 weakly suppresses
  18. L5:875 weakly suppresses
  19. L10:2287 weakly excites
  20. L16:522 weakly suppresses
  21. L15:87 weakly suppresses
  22. L26:510 weakly excites
  23. L18:1021 weakly suppresses
  24. L9:2611 weakly suppresses 'is' right after 'there'

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 -1.77 and -1.691 values between -1.53 and -1.461 values between -1.38 and -1.301 values between -1.30 and -1.221 values between -1.22 and -1.141 values between -1.14 and -1.061 values between -0.98 and -0.901 values between -0.90 and -0.831 values between -0.83 and -0.751 values between -0.75 and -0.671 values between -0.67 and -0.591 values between -0.59 and -0.512 values between -0.51 and -0.439 values between -0.43 and -0.3529 values between -0.35 and -0.27130 values between -0.27 and -0.20374 values between -0.20 and -0.12837 values between -0.12 and -0.041350 values between -0.04 and 0.04721 values between 0.04 and 0.12314 values between 0.12 and 0.2085 values between 0.20 and 0.2827 values between 0.28 and 0.368 values between 0.36 and 0.444 values between 0.44 and 0.512 values between 0.51 and 0.591 values between 0.67 and 0.751 values between 0.75 and 0.831 values between 0.91 and 0.991 values between 1.30 and 1.38offnegativepositiveoften

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. deb
  2. бор
  3. DE
  4. .deb
  5. icolor
  6. 百家
  7. borrower
  8. ammen
  9. (DE
  10. .BL

Suppresses

  1. uces
  2. =<?=
  3. faces
  4. =min
  5. parameters
  6. ucing
  7. fläche
  8. 碣
  9. ean
  10. apologies