Neuron Explorer

Qwen3-0.6B

operators and symbols in code or formulas

A neuron in layer 14 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

It fires on symbols such as plus signs, slashes, question marks and single letters that sit inside code or maths formulas. It is a mixed bag of technical notation. Turning it down made the model get stuck in repeated brackets.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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

    *) memchr (str, 'p', strlen(str));⏎printf ("'p' found at position %d.\n", pch-str+1);⏎printf ("'p' not found.\n");⏎- Compare two blocks

  2. #2 peak

    unorm = sqrt(u(1)*u(1)+u(2)*u(2)); %calculate intermediate value⏎udot(1) = -u(1)/R + u(2) - un<|endoftext|>Thanks for visiting⏎

  3. #3 peak

    folder into which you want to install the free C++Builder 5 command line tool development system.⏎- From the bin directory of your installation: Add “c:\Borland\Bcc55\bin” to the existing path⏎

  4. #4 peak

    ,⏎- times the derivative of sine of 5x with respect to 5x, that's just going to be cosine of 5x.⏎- If we want to write this in differential form, which is useful

  5. #5 peak

    find 386 pages of resources to learn more about Caddo history and culture, https://scholarworks.sfasu.edu/cgi/viewcontent.cgi?article=4012&context=ita⏎Visit the THC’s Learning

  6. #6 peak

    will need⏎- Calculator with statistical functions⏎Define the bel and decibel. The bel is defined mathematically as LB = log10 (P1/P0) where P1 and P0 are two quanties that are in the

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 Last of the Last of the (The (The (The (The (The (The (The (The (The (The (The (The (The (The (The (The (The (The (The (The

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

    = talk_to_server(parse_xml_document=True)If you read it out loud, the code "reads"

  2. #2 peak

    1.5⏎UICBioE240 problem 1.18⏎Write a code that will follow the equation y = x * (x + x) * x.⏎Mile to Kilometer⏎Given A4 sizes find

  3. #3 peak

    = ['raspberry', 'pomegranite', 'bananana', 'grape', 'loganberry'] print "You are an" , random.choice (verb) , random.choice (fruit)⏎Any Tutorial Videos?<|endoftext|>What

  4. #4 peak

    in a safe manner, forceps need to be clamped on to the umbilical cord which has been cut. This cord, even after it is cut is connected to the woman’s

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. L13:2254 suppresses
  2. L12:1576 excites
  3. L8:74 suppresses
  4. L5:2442 excites
  5. L12:473 excites
  6. L13:1855 suppresses opening parenthesis in citations and metadata
  7. L11:2697 weakly excites
  8. L11:2589 weakly suppresses
  9. L5:3043 weakly suppresses
  10. L9:2116 weakly suppresses
  11. L3:131 weakly suppresses
  12. L13:1395 weakly excites
  13. L12:2094 weakly suppresses
  14. L8:892 weakly suppresses
  15. L13:1691 weakly suppresses
  16. L13:1940 weakly excites
  17. L13:2088 weakly excites
  18. L7:5 weakly excites
  19. L3:2567 weakly excites
  20. L3:1348 weakly excites
  21. L10:239 weakly excites
  22. L2:2076 weakly suppresses
  23. L7:482 weakly excites
  24. L2:1780 weakly suppresses

Feeds later layers

  1. L15:1634 suppresses
  2. L17:783 suppresses
  3. L18:2023 suppresses
  4. L15:1857 suppresses
  5. L16:2306 suppresses
  6. L27:86 suppresses
  7. L26:22 suppresses the 'res' fragment inside words
  8. L18:1132 weakly suppresses
  9. L16:2949 weakly suppresses
  10. L18:671 weakly suppresses
  11. L19:1497 weakly suppresses
  12. L17:1959 weakly suppresses
  13. L16:601 weakly excites
  14. L17:488 weakly suppresses
  15. L16:1723 weakly suppresses
  16. L16:2503 weakly suppresses
  17. L18:115 weakly suppresses
  18. L16:1777 weakly suppresses
  19. L16:1254 weakly suppresses
  20. L15:2491 weakly suppresses
  21. L15:617 weakly suppresses
  22. L17:2861 weakly suppresses
  23. L15:1594 weakly suppresses
  24. L18:862 weakly excites hedges like 'it turns out' and 'if any'

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 -2.50 and -2.421 values between -0.74 and -0.661 values between -0.66 and -0.582 values between -0.58 and -0.5010 values between -0.50 and -0.4225 values between -0.42 and -0.3485 values between -0.34 and -0.26231 values between -0.26 and -0.18449 values between -0.18 and -0.10731 values between -0.10 and -0.02996 values between -0.02 and 0.06692 values between 0.06 and 0.14382 values between 0.14 and 0.22194 values between 0.22 and 0.3074 values between 0.30 and 0.3821 values between 0.38 and 0.4610 values between 0.46 and 0.541 values between 0.54 and 0.622 values between 0.62 and 0.70offnegativepositiveoften

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. :date
  3. ended
  4. woff
  5. :A
  6. deserve
  7. .HOUR
  8. uchs
  9. :end
  10. ;,

Suppresses

  1. ()}
  2. {}{⏎
  3. concat
  4. ']}}</
  5. lient
  6. lian
  7. чь
  8. Guys
  9. 揮
  10. li