Neuron Explorer

Qwen3-0.6B

middle pieces of split words

A neuron in layer 0 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 a fragment in the middle of a word that the tokenizer has cut into pieces, such as 'pos' in 'positional', 'ell' in 'propellants' or 'L' in 'Montpelier'. It goes strongly negative on the 'state' in 'prostate'. Turning it up or down did not give a clear change in the writing.

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

    itional and therefore subjective.⏎Propositional – deal with verifiable facts. Either true or false. We can legitimately ask ‘is it true?’⏎Non-propositional statements are those which are subjective. They cannot be

  2. #2 peak

    2014|⏎|ISBN: ||9781312744967 (DRM-EPUB)|⏎|Read Aloud: ||not allowed|<|endoftext|>In 19

  3. #3 peak

    PDF ebook Tiny Bird by Robert Burleigh read. MP3 book Tiny Bird Robert Burleigh buy on iPad.⏎FB2 ebook Tiny Bird on IndieBound. EPUB Tiny Bird by Robert Burleigh for Mac. MOBI book Tiny Bird buy cheap

  4. #4 peak

    .⏎Ocean biology provides many unique examples of life cycles, adaptations and important relationships among organisms (symbiosis, predator-prey dynamics and energy transfer) that do not occur on land.⏎Fundamental Concept

  5. #5 peak

    ⏎Give the values of y when x=3 and when x=n.⏎SOURCE: Carpenter, Corbitt, Kepner, Lindquist, and Reys, 1981. Used by permission of the National Council of Teachers

  6. #6 peak

    's youthful enthusiasm, passed the hat and collected $10.00.⏎Flush with funds, Boylan and Cook experimented with various rocket nozzle designs and propellants. Within months, they successfully launched a 4-foot rocket, with Boy

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 have been thinking about how to get to the point to the point of my life. I have been thinking about how to get to the point to the point of my life. I have been thinking about how to get to the point to

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

    (prostate cancer), spine / spinal cord (metastases, spinal neurinomas,

  2. #2 peak

    samples used in this protocol were from patients who underwent radical prostatectomy for clinically localized prostate cancer; however other tissues and cell lines can be substituted in. Prostate specimens were snap-frozen in liquid nitrogen after resection. Clinical variables and follow

  3. #3 peak

    some point in their life.⏎However, and amazingly true, Prostate cancer is one of the most treatable forms of cancer — if it is diagnosed

  4. #4 peak

    that are specific to lung cancer. This unique skill has the potential to revolutionize the early detection of lung cancer and improve patient outcomes.⏎1.3. Prostate cancer⏎Prostate cancer is the most common cancer among men. It affects the

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

None: this is the first layer.

Feeds later layers

  1. L22:1794 weakly excites "shown in" before a figure reference
  2. L17:2735 weakly excites
  3. L21:3049 weakly excites
  4. L23:2622 weakly excites
  5. L26:2319 weakly excites
  6. L9:2407 weakly suppresses
  7. L14:1949 weakly suppresses
  8. L25:1639 weakly excites
  9. L10:1651 weakly suppresses
  10. L18:1647 weakly excites
  11. L12:2876 weakly excites
  12. L17:2607 weakly excites
  13. L19:514 weakly suppresses
  14. L2:1175 weakly suppresses
  15. L17:1077 weakly excites
  16. L6:1067 weakly excites
  17. L22:2530 weakly excites
  18. L13:2428 weakly suppresses
  19. L13:2026 weakly suppresses
  20. L4:545 weakly suppresses
  21. L10:488 weakly excites
  22. L27:2112 weakly excites
  23. L2:335 weakly suppresses
  24. L21:1272 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 -2.05 and -1.981 values between -0.77 and -0.701 values between -0.63 and -0.571 values between -0.57 and -0.501 values between -0.50 and -0.431 values between -0.43 and -0.367 values between -0.36 and -0.3010 values between -0.30 and -0.2346 values between -0.23 and -0.16181 values between -0.16 and -0.09611 values between -0.09 and -0.032312 values between -0.03 and 0.04567 values between 0.04 and 0.11126 values between 0.11 and 0.1831 values between 0.18 and 0.247 values between 0.24 and 0.312 values between 0.45 and 0.511 values between 0.51 and 0.581 values between 0.58 and 0.65offnegativepositiveoften

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. isée
  2. istique
  3. imedia
  4. wins
  5. ,w
  6. ège
  7. cedes
  8. иск
  9. øy
  10. ickle

Suppresses

  1. throat
  2. 喉咙
  3. 目の
  4. <header
  5. <tbody
  6. 尽可能
  7. bul
  8. ebp
  9. _PUR
  10. <Renderer