Neuron Explorer

Qwen3-0.6B

web addresses and links, URL parts

A neuron in layer 1 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on the parts of web addresses and link text, such as the spot right after a site name or the word before a link. When it was pushed up, the model's writing wandered into news-style and everyday topics, but the link pattern itself was not clearly carried over.

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

    This exodus is known as the Great Migration, and was the first phase of an African American migration that would continue until 1970.⏎Additional resources for research⏎- The Great Migration, In

  2. #2 peak

    control for situations that the machine can’t quite do things,” he said.⏎For the rest of this article, click here: http://www.abc.net.au/news/rural/2017-05-23/the-future

  3. #3 peak

    in business.⏎This week’s article by Jeremy Rosen on Rav Lichtenstein is available from http://jeremyrosen.blogspot.co.uk/2015/08/rav-lichtenstein-zl.html⏎Your comment will be posted

  4. #4 peak

    CA.⏎Other Resources on Rumex acetosella⏎- USDA PLANTS database - Federal database with information

  5. #5 peak

    lights, laminar flow cabinets, and other biological safety cabinet components protect.⏎Read More @ https://cmibloginsight.blogspot.com/2023/05/biological-safety-cabinets-ensuring.html<|endoftext|>World

  6. #6 peak

    News Erutpion states it as being restless. http://volcansvanuatueruptionsgb.blogspot.fr/2014/03/08032014-nevado-de-colima-cleveland

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 am thinking about lately: I am thinking about lately: I am thinking about lately: I am thinking about lately: I am thinking about lately: I am thinking about lately: I am thinking about lately: I am thinking about lately:

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

    0 us, https://commons.wikimedia.org/w/index.php?curid=12708190<|endoftext|>Ayurvedic medicine

  2. #2 peak

    structure = function.⏎(Image from: https://commons.wikimedia.org/w/index.php?curid=1267900)⏎Want to suggest an edit

  3. #3 peak

    already available on the CRIN website but might not be easily retrieveable.⏎See the Wiki here: http://wiki.crin.org/mediawiki/index.php?title=Main_Page⏎We are launching the Wiki with an initial batch of

  4. #4 peak

    ⏎3rd sing masc – lu⏎3rd mass masc – lo⏎see also: http://pidweb.ii.uam.es/coser/contenido.php?es&publicaciones⏎South-Central Italian (Rohlfs, 1

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. L0:1301 suppresses
  2. L0:1224 suppresses
  3. L0:178 suppresses
  4. L0:1 suppresses
  5. L0:2587 weakly suppresses
  6. L0:799 weakly suppresses
  7. L0:1369 weakly suppresses
  8. L0:2507 weakly excites
  9. L0:188 weakly excites
  10. L0:4 weakly suppresses
  11. L0:2500 weakly suppresses
  12. L0:639 weakly excites
  13. L0:537 weakly suppresses the 'or' in 'and/or'
  14. L0:798 weakly excites
  15. L0:1697 weakly suppresses
  16. L0:60 weakly excites
  17. L0:539 weakly suppresses
  18. L0:3012 weakly excites
  19. L0:296 weakly suppresses
  20. L0:1121 weakly excites
  21. L0:93 weakly suppresses
  22. L0:1249 weakly suppresses
  23. L0:1536 weakly excites
  24. L0:2468 weakly excites

Feeds later layers

  1. L18:1979 suppresses
  2. L5:3014 excites
  3. L16:2595 weakly suppresses
  4. L10:2469 weakly excites
  5. L4:1957 weakly excites
  6. L14:1104 weakly suppresses
  7. L16:1124 weakly suppresses
  8. L9:2857 weakly excites
  9. L6:1752 weakly excites
  10. L19:1767 weakly suppresses
  11. L5:391 weakly suppresses
  12. L18:2749 weakly suppresses
  13. L21:1284 weakly excites
  14. L15:2095 weakly suppresses
  15. L14:1015 weakly suppresses
  16. L21:2166 weakly suppresses
  17. L8:133 weakly excites
  18. L26:2960 weakly excites
  19. L6:770 weakly excites
  20. L9:2301 weakly excites
  21. L6:231 weakly suppresses
  22. L2:1251 weakly excites
  23. L3:2965 weakly excites
  24. L17:1008 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.

3 values between -0.29 and -0.273 values between -0.27 and -0.252 values between -0.25 and -0.236 values between -0.23 and -0.2112 values between -0.21 and -0.189 values between -0.18 and -0.1625 values between -0.16 and -0.1434 values between -0.14 and -0.1257 values between -0.12 and -0.1067 values between -0.10 and -0.08156 values between -0.08 and -0.05383 values between -0.05 and -0.03957 values between -0.03 and -0.011386 values between -0.01 and 0.01495 values between 0.01 and 0.03157 values between 0.03 and 0.0661 values between 0.06 and 0.0837 values between 0.08 and 0.1017 values between 0.10 and 0.1218 values between 0.12 and 0.145 values between 0.14 and 0.178 values between 0.17 and 0.193 values between 0.19 and 0.211 values between 0.21 and 0.231 values between 0.23 and 0.251 values between 0.30 and 0.322 values between 0.36 and 0.381 values between 0.41 and 0.431 values between 0.56 and 0.58offnegativepositiveoften

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. 读后
  2. 寤
  3. 解
  4. 业
  5. 三次
  6. clearTimeout
  7. 楔
  8. 两次
  9. 熄
  10. ████

Suppresses

  1. quelle
  2. gue
  3. otide
  4. sense
  5. viewpoint
  6. egas
  7. →statement
  8. …”
  9. otype
  10. ätze