Neuron Explorer

Qwen3-0.6B

unclear: mixed mid-sentence words

A neuron in layer 15 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 grab bag of single words in the middle of sentences, such as 'pronounced', 'saw', 'observed' and 'TITLE', with no clear shared meaning. When it was turned up, the model wrote rambling, repetitive text, so it does not seem to control one clear topic.

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

    entire universe, US-based Space Foundation CEO Tom Zelibor had said in April.⏎On April 12, 1961, Gagarin pronounced his famous “Poyekhali!” (Let’s Go!) as the

  2. #2 peak

    data on on sicklefins, tagged them, and collected sperm and eggs for captive rearing. For more photos, visit: https://www.flickr.com/photos/usfwssoutheast/albums/72157667

  3. #3 peak

    civil liberties in an age of terrorism.⏎5. Justify their stances on this topic in letters to state legislators.⏎Resources / Materials:⏎-five index cards with instructions, as

  4. #4 peak

    an escape character, an escape is a single backward slash ( \ ) in Linux, programming, and regular expressions that perform a special function or to

  5. #5 peak

    War I. Yet the son declined to disappear from the scene and played roles both dramatic and useful over the eight decades of his maturity.⏎He worried Hitler, who saw him as a potential threat to the Anschluss uniting Austria with Germany

  6. #6 peak

    religious ceremony, in South Dakota this past summer.⏎Complete information about survivors could not be confirmed<|endoftext|>A simple example of HTML is:⏎<HTML> <HEAD> <TITLE>This is the Title</TITLE> </HEAD> <BODY

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: Let me look at this, I will make an argument, assume that I will assume the following, Let me assume that: "Suppose the following, let me make the assumption: " Let me make the argument

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

    -reviewed articles.⏎Write a 7–8-page paper in Word format. Apply APA standards to citation of sources. Use the following file naming convention: LastnameFirstInitial_M5_A1.doc.⏎By Saturday, August 16,

  2. #2 peak

    Kansas’s support for the Federals “should be broken,” he would not “destroy that which is absolutely necessary for the subsistence–I may almost say the existence–of the surrounding inhabitants.” Price proposed destroying the Hannibal & St. Joseph

  3. #3 peak

    gaps in order to provide students with instruction. The problem is not that schools do not wish to hire certified teachers but that a suitable pool of certified teachers qualified to teach in these fields and willing to work in these districts does not exist. In addition

  4. #4 peak

    in Jewish communities around the world today.⏎|Extinct||Mishnaic Hebrew extinct as a spoken language by the 5th century CE, surviving as a liturgical language along with Biblical Hebrew for Judaism|⏎Is Arabic easier than Hebrew?⏎

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. L13:687 suppresses
  2. L12:766 excites
  3. L12:2559 suppresses
  4. L14:2207 excites
  5. L14:2126 excites
  6. L14:974 excites
  7. L14:2973 excites
  8. L11:823 suppresses
  9. L14:714 weakly excites
  10. L13:2198 weakly excites
  11. L14:2358 weakly excites
  12. L14:889 weakly excites
  13. L12:603 weakly excites
  14. L14:1619 weakly excites
  15. L12:917 weakly excites
  16. L12:785 weakly excites
  17. L0:2696 weakly suppresses
  18. L14:2352 weakly suppresses
  19. L13:2384 weakly excites
  20. L14:902 weakly excites
  21. L6:2939 weakly suppresses
  22. L1:1681 weakly suppresses
  23. L13:2140 weakly suppresses
  24. L14:1789 weakly suppresses

Feeds later layers

  1. L16:1871 excites
  2. L16:984 suppresses
  3. L16:335 suppresses
  4. L16:2757 suppresses
  5. L20:697 weakly excites
  6. L16:2029 weakly suppresses
  7. L16:1622 weakly suppresses
  8. L16:252 weakly suppresses
  9. L16:291 weakly excites
  10. L16:716 weakly excites
  11. L17:1912 weakly suppresses
  12. L16:2762 weakly excites
  13. L18:2307 weakly excites
  14. L17:1788 weakly suppresses
  15. L16:593 weakly excites
  16. L21:1439 weakly suppresses
  17. L18:683 weakly excites 'is' after a long subject phrase
  18. L17:490 weakly excites
  19. L19:886 weakly excites the space before a rounded number
  20. L16:884 weakly suppresses
  21. L20:604 weakly excites
  22. L16:909 weakly excites
  23. L16:1663 weakly suppresses
  24. L16:2014 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 -2.66 and -2.541 values between -2.09 and -1.982 values between -1.64 and -1.535 values between -1.53 and -1.423 values between -1.42 and -1.317 values between -1.31 and -1.198 values between -1.19 and -1.084 values between -1.08 and -0.977 values between -0.97 and -0.8616 values between -0.86 and -0.7426 values between -0.74 and -0.6353 values between -0.63 and -0.5282 values between -0.52 and -0.41145 values between -0.41 and -0.29281 values between -0.29 and -0.18611 values between -0.18 and -0.071753 values between -0.07 and 0.05533 values between 0.05 and 0.16195 values between 0.16 and 0.2777 values between 0.27 and 0.3835 values between 0.38 and 0.5017 values between 0.50 and 0.6113 values between 0.61 and 0.7212 values between 0.72 and 0.836 values between 0.83 and 0.956 values between 0.95 and 1.062 values between 1.06 and 1.171 values between 1.17 and 1.281 values between 1.28 and 1.402 values between 1.51 and 1.621 values between 1.62 and 1.732 values between 1.73 and 1.85offnegativepositiveoften

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. BeNull
  2. ContentSize
  3. apphire
  4. .CommandType
  5. ifest
  6. _features
  7. kaar
  8. remainder
  9. anything
  10. elementType

Suppresses

  1. ',{
  2. _SERIAL
  3. MX
  4. ANTE
  5. ROLL
  6. MC
  7. sucht
  8. ective
  9. ditor
  10. 调配