Neuron Explorer

Qwen3-0.6B

colon after 'Published' in bylines

A neuron in layer 8 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 colon right after the word 'Published' in article headers, such as 'Published: September 2020'. When it is turned up, the text does not clearly change in a way that matches this.

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

    four million, which would imply that reaching the fourth million occurred in less than thirty years.⏎Autor: gf Ana L. Dávila Román⏎Published: September 20,

  2. #2 peak

    could be challenged.⏎"British culture was extremely strait-laced in the 1950s<|endoftext|>By ANAHAD O'CONNOR⏎Published: November 21, 2006⏎THE CLAIM -- Pricking

  3. #3 peak

    Science Grants by Hughes Institute⏎By WILLIAM CELIS 3d⏎Published: October 4, 1994⏎An $86 million grant

  4. #4 peak

    famous Krimml waterfalls; as well as 551 Alpine lakes between <|endoftext|>Gonorrhoea - the drugs don't work⏎Published: 23rd Dec 2011 08:41

  5. #5 peak

    that policies mandating free water access in schools may help to improve excellence in access, and more work is needed to help all schools excel in this area.⏎Published: September 2020⏎ID #: 73248⏎

  6. #6 peak

    0.3390/ijerph9114170⏎Published: 15 November 2012⏎Abstract: Understanding the effects of

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 a new blog at: ... and so on: ... So, for example, I have a new blog at: ... The problem is, I have a new one: So, I have a new one: The

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

    GNU Oflox⏎This is a GNU package:octave⏎released on 31 May 2012⏎|License||Verified by||Verified on||Notes|⏎|GPLv2||Kelly Hopkins||6 December

  2. #2 peak

    1939 – 4 June 1948⏎|Preceded by||James Barry Munnik Hertzog|⏎|Succeeded by||Daniel François Malan|⏎3 September 1919 – 3

  3. #3 peak

    across the US, this is worth a read. The full article can be accessed here.⏎Tags: New York Times, Insect Population Decline<|endoftext|>First posted on Myth & Moor in 2015.⏎"In Britain there were once hundreds

  4. #4 peak

    .-H. (2014, July). Effectiveness of a tiered intervention model in Chinese: A growth model of reading fluency. Paper presented at the Twenty-First Annual Meeting: Society for the Scientific Study of Reading, Santa

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. L7:52 excites
  2. L7:1784 weakly excites punctuation right after 'of' or 'with'
  3. L6:12 weakly excites
  4. L7:547 weakly suppresses
  5. L0:3005 weakly excites
  6. L4:568 weakly excites
  7. L4:3026 weakly suppresses
  8. L5:2005 weakly excites
  9. L7:2074 weakly excites
  10. L7:162 weakly suppresses
  11. L5:94 weakly suppresses
  12. L7:3011 weakly excites
  13. L5:96 weakly suppresses
  14. L0:1023 weakly excites
  15. L7:2204 weakly suppresses
  16. L6:1136 weakly excites
  17. L6:18 weakly excites
  18. L5:384 weakly excites
  19. L4:378 weakly suppresses
  20. L7:1266 weakly suppresses
  21. L3:514 weakly suppresses
  22. L1:2683 weakly excites
  23. L3:187 weakly excites
  24. L7:191 weakly excites

Feeds later layers

  1. L9:2417 suppresses
  2. L9:1724 suppresses
  3. L9:326 weakly suppresses
  4. L10:125 weakly suppresses
  5. L11:58 weakly excites web address colon in urls
  6. L23:2183 weakly excites
  7. L14:1292 weakly suppresses
  8. L22:2255 weakly suppresses
  9. L16:2833 weakly suppresses
  10. L13:2993 weakly suppresses
  11. L10:2570 weakly suppresses
  12. L10:105 weakly excites
  13. L27:1097 weakly excites
  14. L25:1304 weakly excites
  15. L19:2409 weakly suppresses
  16. L23:676 weakly excites the 'm' in a.m. and p.m
  17. L22:2907 weakly suppresses
  18. L14:710 weakly suppresses
  19. L9:82 weakly suppresses
  20. L19:2327 weakly suppresses
  21. L10:2583 weakly suppresses
  22. L9:270 weakly suppresses
  23. L26:340 weakly suppresses
  24. L22:361 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 -6.79 and -6.611 values between -1.16 and -0.974 values between -0.59 and -0.4087 values between -0.40 and -0.221092 values between -0.22 and -0.032135 values between -0.03 and 0.16529 values between 0.16 and 0.3556 values between 0.35 and 0.543 values between 0.54 and 0.72offnegativepositiveoften

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. gemäß
  2. //}⏎⏎
  3. //}⏎
  4. mute
  5. icers
  6. .Misc
  7. nuts
  8. вали
  9. niest
  10. comparable

Suppresses

  1. :@
  2. ensor
  3. suming
  4. 關係
  5. otion
  6. 濡
  7. à
  8. عنه
  9. :
  10. 传感器