Neuron Explorer

Qwen3-0.6B

'by' after a title or author credit

A neuron in layer 6 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 word 'by' when it introduces an author or creator, as in 'articles by', 'poems by', 'translated by' or 'printed from'. Turning it up produced repetitive, unrelated text.

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

    Nursing Research, 24(2), 195-216. My approach is based, in part, on the trailblazing article by Bettina Bergmann, The Roman House as Memory Theater.”B. Bergmann

  2. #2 peak

    Representation considers the ways in which ethnic and cultural communities have been represented in NFB documentaries. In it, you will find film clips, archival materials, original articles by NFB experts, interviews with filmmakers and a classroom activity.<|endoftext|>HYDERAB

  3. #3 peak

    , detail. Artwork and poems by migrants tell their own stories in expressively rich visual and verbal images.⏎Research reports

  4. #4 peak

    Eye, one of the most popular viewpoints for tourists.⏎Under the British Empire, which stretched out<|endoftext|>(excerpt from “A Study of Freeway Traffic Operation” by Charles J. Keese, Charles Pinnell

  5. #5 peak

    from the Seven Seas” by Stan Hugill (1st ed p 219).⏎The lyrics: “Shanties from the Seven Seas” by Stan Hugill (1st ed p 219, 2

  6. #6 peak

    use the same architecture,<|endoftext|>A Hundred Verses from Old Japan (The Hyakunin-isshu), tr. by William N. Porter, , at sacred-texts.com⏎LONG is the mountain

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: this is a test question. The question is: "I have been thinking about lately." So, the question is: "This is a test question." The question is: "I have been thinking about lately." So, the question is

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

    business in the world, and energy is its newsletter. Every Tuesday and Thursday, Energy Source will be sent directly to your inbox, bringing you important news, forward-looking analysis and insider intelligence. Register here.⏎According to data from the U.S.

  2. #2 peak

    am is ideal. False Dragonhead can tolerate both full sun and partial shade.<|endoftext|>Copyright © University of Cambridge. All rights reserved.⏎'The Solid' printed from http://nrich.maths.org/⏎is one of a series of problems designed

  3. #3 peak

    um are available?⏎Dried chrysanthemum flowers can be found<|endoftext|>Copyright © University of Cambridge. All rights reserved.⏎'Trick or Treat' printed from http://nrich.maths.org/⏎Emily and Tristan from St James' School

  4. #4 peak

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

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. L5:1621 suppresses
  2. L5:1468 weakly excites
  3. L5:757 weakly excites
  4. L4:2310 weakly excites
  5. L4:565 weakly excites
  6. L3:630 weakly excites
  7. L1:2382 weakly excites
  8. L0:2933 weakly suppresses
  9. L2:192 weakly excites
  10. L4:1623 weakly excites
  11. L5:247 weakly suppresses
  12. L5:2817 weakly excites
  13. L1:1269 weakly excites
  14. L1:1464 weakly excites
  15. L3:1441 weakly suppresses
  16. L5:1997 weakly excites
  17. L4:2781 weakly suppresses
  18. L4:2917 weakly excites
  19. L4:1839 weakly excites
  20. L5:157 weakly excites
  21. L3:2672 weakly suppresses
  22. L2:2208 weakly excites
  23. L1:1335 weakly excites
  24. L5:771 weakly excites

Feeds later layers

  1. L9:1627 excites
  2. L9:1562 excites 'by' before a byline or agent
  3. L14:1796 excites
  4. L10:223 excites
  5. L15:2606 excites
  6. L7:2571 excites
  7. L10:629 excites
  8. L11:2427 excites
  9. L13:2465 excites
  10. L14:2508 excites
  11. L12:1286 excites
  12. L9:1830 excites
  13. L15:2889 excites
  14. L16:2208 excites
  15. L22:399 excites
  16. L11:361 excites
  17. L15:2005 excites
  18. L21:772 excites
  19. L7:304 excites
  20. L18:56 excites
  21. L10:3018 excites
  22. L9:1961 excites 'of' in titles and place names
  23. L10:1054 weakly excites 'from' after movement or transfer verbs
  24. L19:2006 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 -0.95 and -0.863 values between -0.68 and -0.596 values between -0.59 and -0.5016 values between -0.50 and -0.4145 values between -0.41 and -0.32108 values between -0.32 and -0.23276 values between -0.23 and -0.13652 values between -0.13 and -0.041201 values between -0.04 and 0.05824 values between 0.05 and 0.14472 values between 0.14 and 0.23190 values between 0.23 and 0.3260 values between 0.32 and 0.4125 values between 0.41 and 0.5010 values between 0.50 and 0.594 values between 0.59 and 0.682 values between 0.68 and 0.774 values between 0.77 and 0.862 values between 1.13 and 1.223 values between 1.40 and 1.491 values between 1.67 and 1.761 values between 1.76 and 1.852 values between 2.57 and 2.66offnegativepositiveoften

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. Amen
  2. isc
  3. ôme
  4. aine
  5. commissioners
  6. ACM
  7. Commissioners
  8. ån
  9. 年代
  10. 层

Suppresses

  1. 彀
  2. PAD
  3. 直
  4. ��
  5. emap
  6. >Data
  7. otional
  8. �
  9. 过得
  10. nqu