Neuron Explorer

Qwen3-0.6B

'://' in web addresses

A neuron in layer 14 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 '://' in web addresses, like the one after 'http' or 'https'. Turning it up did not make the writing about links or web pages, so its effect on output is unclear.

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

    |Contact details of provider:|| Web page: http://www.tandfonline.com/VECE20|⏎|Order Information:||Web: http://www.tandfonline.com/pricing/journal/VECE20|⏎

  2. #2 peak

    groundwater for hydro electric power generation. Journal of Hydrology (NZ), 34(1), 1-14.⏎Permanent Research Commons link: http://hdl.handle.net/10289/4701⏎Control

  3. #3 peak

    :<|endoftext|>Copyright © University of Cambridge. All rights reserved.⏎'Aim High' printed from http://nrich.maths.org/⏎Why do this problem?⏎offers the opportunity for a

  4. #4 peak

    >.⏎- APA: Learning Math; Measurement: Indirect Measurement And Trigonometry. Boston, MA: WGBH Media Library & Archives. Retrieved from http://openvault.wgbh.org/catalog/V_4E4EE3EAD

  5. #5 peak

    , vol.25, n.1, pp. 44-69. ISSN 1980-4369. http://dx.doi.org/10.1590/S010

  6. #6 peak

    Kansas Sand Hills,"⏎Proceedings of the Iowa Academy of Science, 31(1), 131-132.⏎Available at: https://scholarworks.uni.edu/pias/vol31/iss1/2

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 to be able to see this. I...? I think I am going to try to do it? ... I feel the way I do now, and I am trying? I am going to try. I think I have

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

    1100.00010000.1yyyyyyy.yyyyyyyy => 172.16.128.

  2. #2 peak

    of white-nosed coati (9.05%), being this difference significant in summer 2012 (P < 0.00001), summer (P =

  3. #3 peak

    , though, and there’s a lengthy list of variations still kicking around in the wild, such as ay-ight, aiite, and even the occasional aaaaaaaight. As the dictionary indicates, aight’s colloquial English,

  4. #4 peak

    file should contain one line with one integer: the largest diameter of any shape in the painting.⏎4 9 XXXX.XXXX ..XX.X... XX.XXX..X ....X..XX⏎There are three shapes in this example, labelled

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. L10:1597 weakly suppresses
  2. L7:2428 weakly suppresses
  3. L12:401 weakly excites
  4. L0:1224 weakly suppresses
  5. L4:55 weakly suppresses
  6. L8:1077 weakly suppresses
  7. L6:2333 weakly excites
  8. L4:2639 weakly suppresses
  9. L10:3065 weakly excites
  10. L6:1188 weakly excites
  11. L10:1165 weakly excites
  12. L8:4 weakly excites
  13. L2:4 weakly suppresses
  14. L9:33 weakly suppresses
  15. L9:87 weakly suppresses
  16. L7:238 weakly excites
  17. L8:1308 weakly suppresses
  18. L10:2219 weakly suppresses
  19. L11:2 weakly suppresses unclear: blank before a year or number
  20. L11:2172 weakly suppresses
  21. L4:382 weakly suppresses
  22. L5:309 weakly suppresses
  23. L7:2581 weakly suppresses
  24. L3:1765 weakly excites

Feeds later layers

  1. L15:760 weakly excites
  2. L15:1404 weakly excites
  3. L15:2444 weakly excites
  4. L22:882 weakly excites
  5. L17:2896 weakly suppresses
  6. L19:757 weakly suppresses
  7. L24:908 weakly excites
  8. L17:1947 weakly suppresses
  9. L23:1978 weakly excites
  10. L24:327 weakly excites
  11. L24:181 weakly suppresses
  12. L17:2748 weakly excites
  13. L22:2244 weakly excites
  14. L21:771 weakly excites
  15. L26:116 weakly suppresses
  16. L20:858 weakly excites
  17. L20:504 weakly excites
  18. L21:1444 weakly excites
  19. L27:101 weakly suppresses
  20. L20:1816 weakly excites
  21. L17:1024 weakly suppresses
  22. L21:497 weakly excites
  23. L21:2598 weakly excites
  24. L27:923 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 -1.04 and -0.991 values between -0.99 and -0.954 values between -0.64 and -0.592 values between -0.59 and -0.553 values between -0.55 and -0.5010 values between -0.50 and -0.4622 values between -0.46 and -0.4240 values between -0.42 and -0.3769 values between -0.37 and -0.3381 values between -0.33 and -0.28171 values between -0.28 and -0.24237 values between -0.24 and -0.19280 values between -0.19 and -0.15350 values between -0.15 and -0.10440 values between -0.10 and -0.06476 values between -0.06 and -0.02541 values between -0.02 and 0.03359 values between 0.03 and 0.07284 values between 0.07 and 0.12190 values between 0.12 and 0.16134 values between 0.16 and 0.2182 values between 0.21 and 0.2562 values between 0.25 and 0.3025 values between 0.30 and 0.3415 values between 0.34 and 0.3910 values between 0.39 and 0.434 values between 0.43 and 0.475 values between 0.47 and 0.525 values between 0.52 and 0.561 values between 0.56 and 0.611 values between 0.61 and 0.652 values between 0.65 and 0.701 values between 0.70 and 0.74offnegativepositiveoften

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. indre
  2. ходим
  3. ifax
  4. ches
  5. enf
  6. печ
  7. otify
  8. icie
  9. ienie
  10. opic

Suppresses

  1. by
  2. ManyToOne
  3. تحمل
  4. .Offset
  5. 源头
  6. _BOX
  7. opposition
  8. bara
  9. types
  10. pipelines