Neuron Explorer

Qwen3-0.6B

comma between author names in citations

A neuron in layer 19 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 reacts to the comma that separates an author's surname from their first initials in reference lists, such as 'Davis, L.M., Spackman, E.'. Its negative side is weak and unrelated. Turning it up or down gave text that did not clearly relate to citations.

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

    , 2008⏎Publication Date: May 16, 2008⏎Citation: Davis, L.M., Spackman, E. 2008. Do crocodilians get the flu?

  2. #2 peak

    Keywords: viral marketing, word-of-mouth marketing, consumption patterns, information traits, credibility.⏎ANTHEUNIS, M.L., & REIJMERSDAL, E.A. (2012). Social connections and the persuasiveness

  3. #3 peak

    Whole System Simulation, Proceedings of the 13th USENIX Security Symposium, 2004.⏎Garfinkel, S. and Shelat, A., "Remembrance of Data Passed: A Study of Disk Sanitization

  4. #4 peak

    an resistance to the host.⏎References[change | change source]⏎- Callmigration.org: Gene targeting⏎- Jang, Chuan-Wei; Magnuson, Terry (2013). "A novel selection marker for efficient DNA cloning

  5. #5 peak

    week for signs of a white<|endoftext|>Disease induced changes in gene flow patterns among Tasmanian devil populations⏎Bruniche-Olsen, A and Burridge, CP and Austin, JJ and Jones, M, Disease induced changes in gene flow

  6. #6 peak

    ||Endangered B1ab(i,ii,iii,v) ver 3.1|⏎|Reviewer(s):||Butchart, S. & Symes, A.|⏎|Contrib

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 I have been I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I 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

    set of equations.⏎y = 2x +3 and 4y = –2x+10⏎On comparing y = 2x +3 and 4y = –2x+10 with y = mx + c

  2. #2 peak

    - Statment 1⏎- Statement 2⏎Regex for checking if its a valid Either of Two words i.e., Either-OR⏎Test string examples for the above regex-⏎|Input String||Match Output|⏎|cho||does not

  3. #3 peak

    ability⏎U.S. Department of Housing and Urban Development⏎The views expressed in this article are those of the author and do not represent the official positions or policies of

  4. #4 peak

    413967-1-768x146-1-300x57.png 0 0 admin https://

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. L18:1325 suppresses
  2. L17:2290 excites
  3. L17:2602 suppresses
  4. L17:1230 suppresses
  5. L16:1334 excites
  6. L18:2091 suppresses
  7. L17:1199 suppresses
  8. L18:1234 suppresses
  9. L18:1602 excites
  10. L17:2287 excites
  11. L16:1063 suppresses
  12. L18:3069 suppresses
  13. L13:2311 suppresses
  14. L18:3026 suppresses
  15. L18:1132 suppresses
  16. L17:1158 excites
  17. L14:1012 excites
  18. L16:2982 excites
  19. L15:1983 excites
  20. L17:171 suppresses
  21. L18:1727 suppresses
  22. L17:1079 excites
  23. L17:2948 excites
  24. L15:1341 suppresses

Feeds later layers

  1. L20:370 excites
  2. L20:439 suppresses
  3. L20:1189 excites
  4. L20:389 suppresses
  5. L21:112 suppresses
  6. L22:90 suppresses
  7. L20:165 suppresses
  8. L22:58 suppresses
  9. L20:1039 excites
  10. L22:99 suppresses
  11. L20:248 suppresses
  12. L20:400 suppresses
  13. L20:60 suppresses
  14. L22:79 suppresses
  15. L22:98 suppresses
  16. L22:84 suppresses
  17. L20:168 suppresses
  18. L22:80 suppresses
  19. L27:470 excites
  20. L22:128 suppresses
  21. L21:404 excites
  22. L22:112 weakly suppresses
  23. L22:73 weakly suppresses
  24. L23:54 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 -1.34 and -1.231 values between -1.23 and -1.121 values between -1.12 and -1.021 values between -1.02 and -0.912 values between -0.91 and -0.807 values between -0.80 and -0.7016 values between -0.70 and -0.5953 values between -0.59 and -0.4897 values between -0.48 and -0.38310 values between -0.38 and -0.27567 values between -0.27 and -0.16893 values between -0.16 and -0.061072 values between -0.06 and 0.05509 values between 0.05 and 0.16204 values between 0.16 and 0.26101 values between 0.26 and 0.3729 values between 0.37 and 0.4817 values between 0.48 and 0.584 values between 0.58 and 0.694 values between 0.69 and 0.803 values between 0.80 and 0.902 values between 0.90 and 1.011 values between 1.01 and 1.122 values between 1.12 and 1.224 values between 1.22 and 1.332 values between 1.33 and 1.441 values between 1.65 and 1.751 values between 1.75 and 1.861 values between 2.07 and 2.181 values between 2.50 and 2.611 values between 2.82 and 2.93offnegativepositiveoften

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. .simple
  2. уй
  3. Core
  4. RAD
  5. rad
  6. textual
  7. Moon
  8. SIM
  9. 旄
  10. owa

Suppresses

  1. ", ⏎
  2. Revenge
  3. ”),
  4. empre
  5. ”⏎⏎
  6. 禀
  7. ahead
  8. ", ⏎
  9. "), ⏎
  10. presso