Neuron Explorer

Qwen3-0.6B

'by' after a past participle

A neuron in layer 1 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 follows a verb like 'represented', 'indicated' or 'referred', as in 'represented by a scatter plot'. Its negative side is weak and fires on 'about' after words like 'story' or 'information'. Turning it up gave text that drifted to new topics, not anything about this word.

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

    worse. This comes as more international analysts have also began to question the accuracy of the Meles government’s double digit economic growth claims and similar disputed government statistics referred by institutions like the<|endoftext|>An hour before liftoff, things were going as planned

  2. #2 peak

    30. It tells you the oil’s thickness or viscosity. A thinner motor oil — indicated by a lower number — flows more easily. Conversely, a thicker oil — with a

  3. #3 peak

    school events each week and time spent watching videos online each week⏎- temperature and watermelon sales⏎Some data is collected for each pair of variables listed and represented by a scatter plot. For each pair of variables, how do the scatter plots support

  4. #4 peak

    room, or humidity produced by water within the capped horticultural tubes that flowers were mounted in during sampling). The intensity of floral humidity produced by flowers, represented by (the average peak difference in relative humidity in the flower species' headspace compared

  5. #5 peak

    10 MPa excess pressure within the magma chamber, generating the stress field. Layers dips by 20-degree increments, with differing elastic moduli, represented by stiffness ratios (EU/EL) ranging from 0.01 to

  6. #6 peak

    use of memes. A meme is a conception, technique or custom that escalates from one person to another with the purpose of sending a particular meaning or idea represented by the meme. The spread of memes has significantly shaped how people communicate, and they

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 a 24-year-old woman, 22 years old, I am in my 23rd month. I am 24 years old, I am 22 years old, and I am in my

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

    . She befriends a little boy called Amir who is very unhappy when it is time to say goodbye. Perhaps Pebble could keep him company too?⏎A story about the power of friendship in desperate circumstances.⏎As Lubna talks to Pebble,

  2. #2 peak

    a writing folder as a<|endoftext|>This page provides information about the questions in Sundanese, also called the interrogative form. This will

  3. #3 peak

    being pro-choice in this song, he talks about being anti-street-harassment in this song. But he also — it's an autobiographical song about being a young black male growing up<|endoftext|>Despite progress, measles kills 40

  4. #4 peak

    insights TikTok is gathering on its users. TikTok app user interaction data can be utilized to create detailed individual “psychometric” profiles. These profiles contain information about a person’s interests, hobbies, emotional triggers, lifestyle choices, etc. Such

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. L0:2257 weakly suppresses
  2. L0:1651 weakly excites
  3. L0:1232 weakly excites
  4. L0:376 weakly excites
  5. L0:1812 weakly excites
  6. L0:2806 weakly suppresses
  7. L0:2845 weakly suppresses
  8. L0:1943 weakly suppresses
  9. L0:839 weakly excites
  10. L0:1830 weakly excites
  11. L0:1035 weakly suppresses
  12. L0:909 weakly suppresses
  13. L0:1861 weakly excites
  14. L0:2447 weakly excites
  15. L0:2007 weakly suppresses 'key' after 'The'
  16. L0:1747 weakly excites
  17. L0:725 weakly excites
  18. L0:1555 weakly excites
  19. L0:421 weakly excites
  20. L0:1979 weakly excites
  21. L0:255 weakly excites
  22. L0:1820 weakly suppresses
  23. L0:285 weakly excites
  24. L0:327 weakly suppresses

Feeds later layers

  1. L8:1097 suppresses
  2. L18:2339 weakly suppresses
  3. L3:355 weakly suppresses
  4. L3:409 weakly suppresses
  5. L15:1676 weakly suppresses punctuation after names and book titles
  6. L9:2067 weakly excites
  7. L11:2124 weakly suppresses
  8. L12:719 weakly excites
  9. L17:2574 weakly suppresses
  10. L21:2377 weakly suppresses
  11. L6:734 weakly suppresses
  12. L8:2322 weakly excites
  13. L5:413 weakly excites
  14. L12:937 weakly suppresses
  15. L3:1772 weakly suppresses
  16. L2:2756 weakly suppresses
  17. L2:610 weakly suppresses
  18. L6:1802 weakly suppresses
  19. L2:26 weakly suppresses
  20. L10:484 weakly suppresses
  21. L9:3005 weakly suppresses
  22. L20:2753 weakly suppresses
  23. L3:804 weakly suppresses
  24. L25:2064 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 -2.56 and -2.451 values between -2.24 and -2.131 values between -1.92 and -1.821 values between -1.82 and -1.712 values between -1.60 and -1.501 values between -1.18 and -1.081 values between -0.97 and -0.873 values between -0.87 and -0.767 values between -0.76 and -0.653 values between -0.65 and -0.5512 values between -0.55 and -0.4413 values between -0.44 and -0.3457 values between -0.34 and -0.23296 values between -0.23 and -0.131228 values between -0.13 and -0.021766 values between -0.02 and 0.09410 values between 0.09 and 0.1968 values between 0.19 and 0.3016 values between 0.30 and 0.406 values between 0.40 and 0.515 values between 0.51 and 0.612 values between 0.61 and 0.721 values between 0.83 and 0.931 values between 1.04 and 1.142 values between 1.25 and 1.351 values between 1.46 and 1.563 values between 1.56 and 1.67offnegativepositiveoften

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. goggles
  2. 랐
  3. 圆满完成
  4. ContextHolder
  5. atical
  6. ophone
  7. collectively
  8. complète
  9. stereotype
  10. 狠抓

Suppresses

  1. nc
  2. entials
  3. xs
  4. STANCE
  5. anja
  6. thic
  7. ning
  8. 多个国家
  9. ![
  10. parcel