Neuron Explorer

Qwen3-0.6B

'of' after 'each' or 'one' or 'all'

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 word 'of' when it follows words like 'each', 'one' or 'all', as in 'each one of these' or 'every one of these studies'. Turning it up gave unrelated text, so the steering does not show this pattern.

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

    c) recognise different rhythms. Watch this lesson first for a recap: Reading Music On The Bass Guitar: The Very Basics - Note Names And Locations.⏎Each one of those things is quite easy to do on its own. At first, when you

  2. #2 peak

    us Mountains. Today, roughly 4,000 Circassians live in Israel, where they constitute the country’s only Sunni Muslim community that sends each of its sons to the military.⏎Read post on Arty Semite blog about Israeli

  3. #3 peak

    , they’re so rare that Peter Ackerman wrote a children’s book, “The Lonely Phone Booth,” about one of only four remaining outdoor phone booths in all of Manhattan.⏎“I walked past this phone booth every day with my kid when he was

  4. #4 peak

    port on the station in order to transfer people and material to sustain a space station.⏎Each one of these feats is meant to fit into the long term objective of having a Chinese manned

  5. #5 peak

    , Roberto Arlt, Alfonsina Storni, and Mário de Andrade, restoring the original newspaper contexts in which their articles first emerged.⏎Each of these writers guided their readers through a constantly changing cityscape and advised them on matters

  6. #6 peak

    as well as the establishment of a strategic military alliance. The British Embassy in Democratic Republic of the Congo maintains and develops relations between the UK and DRC. Each of Francis' projects was luxuriously decorated both inside and out. Francis was also

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: the, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and,

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

    causes may be purely physiological; maybe nothing is particularly wrong in the teen’s life, but his or her brain just seems to run in a depressed mode.⏎What Are the Signs of

  2. #2 peak

    incompatible with the Kolmac treatment program.⏎Relapse Sequence Diagram⏎When a person develops a problem with alcohol or other drugs, one of the consequences is that his or her thinking about these substances becomes distorted in a way that perpetuates the problem.

  3. #3 peak

    your doctor on the alert for symptoms of asbestos–related diseases, which often take decades to develop after the initial asbestos exposure (see asbestosis, mesothelioma, and lung cancer for details).⏎If your asbestos exposure took place during the

  4. #4 peak

    to get some blood pressure readings at home. In some cases, he or she may recommend that you wear an ambulatory blood pressure monitor. This device measures

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. L5:125 strongly suppresses "of" after all three, both, ten
  2. L6:972 strongly suppresses "well" after "how"
  3. L6:1718 strongly excites
  4. L5:1788 strongly excites
  5. L6:1782 strongly suppresses 'of' after a category noun
  6. L6:3016 strongly excites
  7. L1:818 excites
  8. L7:2164 excites period inside country abbreviations
  9. L4:2473 suppresses
  10. L4:1032 suppresses
  11. L6:1195 suppresses
  12. L5:1747 excites
  13. L6:1063 excites
  14. L6:376 suppresses
  15. L6:2785 suppresses
  16. L4:1743 excites
  17. L7:111 excites
  18. L6:2509 suppresses
  19. L6:2327 excites
  20. L4:217 excites
  21. L6:1233 excites
  22. L7:207 suppresses
  23. L6:2928 suppresses
  24. L6:443 excites

Feeds later layers

  1. L10:167 strongly excites
  2. L15:557 strongly suppresses
  3. L17:985 strongly excites
  4. L11:810 strongly suppresses
  5. L19:663 strongly excites
  6. L14:280 strongly suppresses
  7. L14:2271 strongly suppresses
  8. L21:716 strongly excites
  9. L21:274 strongly suppresses
  10. L16:2520 excites
  11. L17:1467 suppresses
  12. L18:602 excites
  13. L20:519 suppresses
  14. L18:217 suppresses
  15. L12:1058 excites
  16. L13:975 excites
  17. L20:337 suppresses
  18. L18:2189 excites
  19. L21:68 suppresses
  20. L18:1089 excites
  21. L18:277 suppresses
  22. L22:360 suppresses
  23. L12:372 excites
  24. L9:820 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 -2.42 and -2.232 values between -2.23 and -2.042 values between -2.04 and -1.843 values between -1.84 and -1.652 values between -1.65 and -1.467 values between -1.46 and -1.2710 values between -1.27 and -1.0819 values between -1.08 and -0.8831 values between -0.88 and -0.6962 values between -0.69 and -0.50141 values between -0.50 and -0.31356 values between -0.31 and -0.121661 values between -0.12 and 0.08878 values between 0.08 and 0.27492 values between 0.27 and 0.46159 values between 0.46 and 0.6536 values between 0.65 and 0.8415 values between 0.84 and 1.048 values between 1.04 and 1.235 values between 1.23 and 1.426 values between 1.42 and 1.614 values between 1.80 and 2.002 values between 2.19 and 2.381 values between 2.95 and 3.151 values between 3.15 and 3.341 values between 3.72 and 3.911 values between 3.91 and 4.111 values between 4.30 and 4.491 values between 5.07 and 5.26offnegativepositiveoften

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. まま
  2. ものが
  3. 特点是
  4. 做法
  5. 内容
  6. ことを
  7. behavior
  8. 样子
  9. 行为
  10. ほうが

Suppresses

  1. ectomy
  2. 的情绪
  3. ously
  4. entially
  5. sobie
  6. otine
  7. ocese
  8. @register
  9. :@"%@
  10. 的因素