Neuron Explorer

Qwen3-0.6B

digits inside citation and DOI strings

A neuron in layer 22 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on the numbers and punctuation inside reference-style strings, like the digits in a journal DOI or a page range in a bibliography. The steering results did not show a clear theme. Turning it up made the text repeat words and phrases, and turning it down gave repeated number patterns.

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

    , 155. 50-58. 10.1016/j.biocon.2012.05.013Full text not available from this repository.⏎Although grasslands are crucial

  2. #2 peak

    -1600. http://dx.doi.org/10.1016/j.jbusres.2013.10.001⏎BERISHA-NAMANI, M. (20

  3. #3 peak

    .⏎"Purchasing Power Parity Exchange Rates for the Global Poor,"⏎American Economic Journal: Applied Economics,⏎American Economic Association, vol. 3(2), pages 137-66, April.⏎- Angus Deaton & Olivier

  4. #4 peak

    13, 5(7), 3006-3008; doi:10.3390/su5073006⏎I am honored to introduce this special issue of Sustainability,

  5. #5 peak

    with Nupecoupé, toupée⏎What do you find interesting about this word or phrase?⏎Comments that don't adhere to our Community Guidelines may be moderated or removed.<|endoftext|>Why travel to Europe to see old ruins when you have some

  6. #6 peak

    Medicine study are available here: http://www.europeangeriaticmedicine.com/article/S1878-7649(13)00123-X/).⏎It is possible nursing staff underestimate the importance of oral

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: in the context of the real-tn-2011-12-24-12-25-12-26-12-27-12-28-12-2

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

    ; CIA, 2003). Furthermore, parks and<|endoftext|>From: Mary Ann Liebert Inc.⏎Posted: Saturday, September 1, 2007⏎Astrobiology August 2007, 7(

  2. #2 peak

    sickly, thin, or emaciated appearance; drawn; said of the face or the expression.⏎from WordNet 3.0 Copyright 2006 by Princeton University. All rights reserved.⏎- adj. somewhat ill or prone

  3. #3 peak

    from Wiktionary, Creative Commons Attribution/Share-Alike License⏎- n. An underwater explorer.⏎from WordNet 3.0 Copyright 2006 by Princeton University. All rights reserved.⏎- n. an underwater swimmer

  4. #4 peak

    two different numbers cant have the same cube). Therefore, by Definition 1, is one-to-one....⏎View Full Document⏎- Spring '09<|endoftext|>part of speech: noun⏎definition 1: an upright line

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. L20:790 excites
  2. L19:1627 excites
  3. L18:2060 suppresses
  4. L19:619 suppresses
  5. L20:1342 excites
  6. L13:1606 excites
  7. L19:1017 weakly excites
  8. L18:764 weakly suppresses
  9. L11:1290 weakly excites
  10. L21:362 weakly excites
  11. L19:1242 weakly excites
  12. L0:730 weakly suppresses
  13. L19:279 weakly excites
  14. L19:2153 weakly suppresses
  15. L20:1090 weakly excites
  16. L14:2170 weakly excites
  17. L20:806 weakly suppresses
  18. L19:616 weakly excites
  19. L4:21 weakly excites
  20. L18:1092 weakly excites
  21. L21:2799 weakly suppresses
  22. L20:3046 weakly excites
  23. L21:2030 weakly excites
  24. L5:364 weakly suppresses

Feeds later layers

  1. L23:363 strongly suppresses
  2. L24:710 suppresses
  3. L25:2507 suppresses
  4. L23:349 excites comma after author surname in citations
  5. L26:386 excites
  6. L23:404 suppresses
  7. L23:2370 suppresses
  8. L26:1756 suppresses
  9. L27:268 suppresses
  10. L23:383 suppresses
  11. L23:1206 weakly excites
  12. L26:219 weakly suppresses
  13. L26:539 weakly suppresses
  14. L25:915 weakly suppresses
  15. L27:1361 weakly suppresses
  16. L27:509 weakly suppresses
  17. L24:2045 weakly suppresses acronym introductions
  18. L23:1690 weakly suppresses
  19. L24:749 weakly suppresses
  20. L27:295 weakly excites
  21. L25:522 weakly suppresses
  22. L27:468 weakly excites
  23. L23:531 weakly suppresses
  24. L25:1223 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 -3.15 and -3.011 values between -2.11 and -1.961 values between -1.66 and -1.514 values between -1.21 and -1.062 values between -1.06 and -0.917 values between -0.91 and -0.7612 values between -0.76 and -0.6118 values between -0.61 and -0.4642 values between -0.46 and -0.31163 values between -0.31 and -0.16617 values between -0.16 and -0.011768 values between -0.01 and 0.14729 values between 0.14 and 0.29294 values between 0.29 and 0.44131 values between 0.44 and 0.5963 values between 0.59 and 0.7325 values between 0.73 and 0.8818 values between 0.88 and 1.0311 values between 1.03 and 1.181 values between 2.68 and 2.83offnegativepositiveoften

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. ớt
  2. Tower
  3. Bellev
  4. Presidents
  5. =v
  6. Mods
  7. лага
  8. `%
  9. 扞
  10. 很正常

Suppresses

  1. Ak
  2. Ci
  3. 堪
  4. Ec
  5. Fn
  6. 兼
  7. Ai
  8. Ai
  9. tn
  10. Mn