Neuron Explorer

Qwen3-0.6B

document boundaries and line ends

A neuron in layer 17 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 where one piece of text seems to end and another begins, such as an end-of-document marker, a line break after a date or a table row, or the end of a sentence before an unrelated heading. Turning it up gave answers that drifted into new, unrelated passages.

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

    , as Haitian, as Colombian, as republicans,<|endoftext|>Telescope Aperture Explained⏎Last Updated: 02 June 2020⏎The aperture is one of the most important characteristics on

  2. #2 peak

    allowed us to propose something that hasn´t yet been possible to study in depth and that now should be taken into serious consideration. Is genomic instability the cause of tum<|endoftext|>The role of the built environment in addressing climate change is clear as buildings account for

  3. #3 peak

    when an online work was consulted is now optional.⏎Selected examples of MLA 7 vs. MLA 8⏎|MLA 7||MLA 8|⏎Article from a library⏎|Web site with no author|⏎The examples above are

  4. #4 peak

    can benefit from signing.⏎"It's easier for them to grasp," said Skaja, an exceptional/gifted student educator with Brevard Public Schools. ...<|endoftext|>When planning your next landscape project, consider using native species. Native plants tend to

  5. #5 peak

    reasons why these results are important is that we know that young people are getting much less sleep than they did just ten years ago," explains Dr Petrovic. "<|endoftext|>This October, celebrate Global Diversity Awareness Month and pay tribute to the diverse minds and

  6. #6 peak

    . After discussion with another friend, Wes Prichard, Prichard suggested to Wennberg<|endoftext|>Home > Preview⏎The flashcards below were created by user⏎on FreezingBlue Flashcards.⏎What factors could affect feet and leg disorders?⏎

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: what is the 24, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1

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

    4, 5] i = len(given_arr) - 1 while(i >= 0): print(given_arr[i]) i = i - 1⏎In this program,⏎- given_arr is the given array.⏎- i is

  2. #2 peak

    the name given to one of the three main components of UDDI, the protocol used to discover Web Services (the other two being Yellow Pages and Green Pages).⏎- abbreviation

  3. #3 peak

    be seen, and the pelvis extends into the ureter⏎- The urinary system consists of two kidneys, two ureters, the bladder and urethra.⏎- The kidney is one of a pair of organs that removes nitrogenous wastes,

  4. #4 peak

    resistance 2⏎- Spread of resistant strains 2.1⏎- Eponym 3⏎- References 4⏎Klebsiella pneumonia tends to affect people with underlying diseases, such as

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. L16:2964 suppresses
  2. L15:1325 excites
  3. L16:186 excites
  4. L15:574 excites
  5. L13:643 excites
  6. L16:1253 excites
  7. L15:1567 suppresses
  8. L15:1313 excites
  9. L16:1826 weakly suppresses
  10. L9:935 weakly excites
  11. L16:496 weakly suppresses
  12. L6:162 weakly suppresses
  13. L8:51 weakly suppresses
  14. L16:1767 weakly suppresses
  15. L14:17 weakly excites
  16. L11:307 weakly excites
  17. L4:1470 weakly suppresses
  18. L0:2711 weakly excites
  19. L7:1322 weakly excites
  20. L15:1075 weakly excites
  21. L7:1763 weakly excites
  22. L14:1045 weakly suppresses
  23. L14:114 weakly excites
  24. L4:1218 weakly suppresses

Feeds later layers

  1. L18:1817 strongly excites
  2. L18:720 excites
  3. L19:1527 excites
  4. L24:96 excites
  5. L18:1059 excites
  6. L19:550 suppresses
  7. L20:1389 excites
  8. L19:975 suppresses
  9. L26:288 suppresses
  10. L26:156 suppresses
  11. L21:1150 suppresses
  12. L19:2310 weakly suppresses
  13. L18:1047 weakly excites
  14. L18:1105 weakly suppresses
  15. L18:1950 weakly suppresses opening parenthesis after a name or title
  16. L22:597 weakly suppresses end of document token
  17. L21:210 weakly suppresses
  18. L18:833 weakly suppresses
  19. L18:1808 weakly suppresses
  20. L22:534 weakly suppresses
  21. L23:194 weakly excites
  22. L18:1338 weakly suppresses
  23. L26:16 weakly excites
  24. L18:1570 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 -3.001 values between -2.69 and -2.541 values between -2.54 and -2.392 values between -2.39 and -2.231 values between -2.08 and -1.931 values between -1.93 and -1.773 values between -1.77 and -1.621 values between -1.62 and -1.472 values between -1.47 and -1.313 values between -1.01 and -0.856 values between -0.85 and -0.705 values between -0.70 and -0.5532 values between -0.55 and -0.39119 values between -0.39 and -0.24451 values between -0.24 and -0.091438 values between -0.09 and 0.071193 values between 0.07 and 0.22461 values between 0.22 and 0.37136 values between 0.37 and 0.5330 values between 0.53 and 0.688 values between 0.68 and 0.841 values between 0.84 and 0.992 values between 0.99 and 1.142 values between 1.14 and 1.303 values between 1.30 and 1.451 values between 1.76 and 1.911 values between 1.91 and 2.061 values between 2.22 and 2.371 values between 2.68 and 2.831 values between 2.83 and 2.98offnegativepositiveoften

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. ++)⏎
  2. ...)⏎
  3. :]⏎
  4. ]⏎
  5. theast
  6. ]⏎
  7. himself
  8. 心意
  9. ...)⏎
  10. ()]⏎

Suppresses

  1. 、⏎⏎
  2. _⏎⏎
  3. /⏎⏎
  4. —⏎⏎
  5. ,⏎⏎
  6. ",⏎⏎
  7. ,⏎⏎
  8. -⏎⏎
  9. •⏎⏎
  10. -⏎⏎