Neuron Explorer

Qwen3-0.6B

unclear: fragments inside split names

A neuron in layer 3 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

The neuron fires on the middle pieces of long or unusual words that are chopped into several parts, such as the 'ust' in 'Procrustean' or the 'o' in 'Robo'. It is strongly turned off by the word 'is' after 'do', and the steering outputs did not show a clear theme.

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

    actually look at the broad measures of aggregate economic activity that define recessions and recoveries, they don’t look like those letters—at best they would be Procrustean fits. For example, the Great Depression of 1929-

  2. #2 peak

    not compelled, as were those whose homes lay north of the Alps, to distort facts before they could make the plants of their native country fit into the procrustean bed of classical descriptions.⏎One of the chief of the commentators and herbalists

  3. #3 peak

    A NASA original, this galaxy is appropriately named for it’s hat-like appearance yet makes us wonder what exactly is going on in it’s center.⏎R Jay Gabany took this shot of the Trific Nebula, which “… illustrates three different types

  4. #4 peak

    balanced breakfast, setting up your metabolism properly for the day. If you have time constraints, consider options such as protein smoothies, healthy muffins, hardboiled eggs or granola with fruit, nuts and seeds.⏎Eliminate or Min<|endoftext|>

  5. #5 peak

    18 species of bat residing in the UK, including the common pipistrelle and greater horseshoe bat.⏎Monitoring bats is vital not just to

  6. #6 peak

    treatment of tuberculosis.⏎Because of the therapeutic climate in Asheville, N.C., he moved there in 1888 to establish the Winyaw Sanitarium for the treatment of diseases of the lungs and throat. Mode

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 a part of the public or private. I have been a part of your life. I have been a part of your life, or my life. I have been a part of your life. I am a part of your

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

    , but as of this writing modules are not given⏎a chance to take action when a new thread is created or destroyed.)⏎However, what extension modules primarily do is to intercede in the HTTP protocol in order to customize how Apache processes and responds

  2. #2 peak

    when he was a kid. All you have to do is find a piece of paper and use your imagination to draw any prehistoric scene you

  3. #3 peak

    learn how disposed brine impacts surface water quality in Pennsylvania⏎The first thing we will do is get a file from HydroShare and upload it into HydroClient. This will allow

  4. #4 peak

    ⏎Will Japan's nuclear crisis end in catastrophe?⏎When the unexpected happens, even at a nuclear power plant, don't freak out. All you have to do is cool the cores. Just add water. If you can't deliver it through a

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. L0:532 weakly suppresses
  2. L2:1576 weakly excites
  3. L0:1456 weakly suppresses
  4. L2:115 weakly excites
  5. L1:534 weakly excites
  6. L0:313 weakly suppresses "and" joining paired nouns
  7. L0:1423 weakly excites
  8. L2:419 weakly suppresses
  9. L1:1519 weakly suppresses
  10. L0:2228 weakly suppresses
  11. L1:2762 weakly excites
  12. L0:1816 weakly excites
  13. L1:1427 weakly excites 'that' after reporting verbs
  14. L1:1681 weakly excites
  15. L1:2616 weakly suppresses
  16. L2:2003 weakly excites 'of' after abstract nouns like 'sense' or 'view'
  17. L2:625 weakly excites
  18. L0:1365 weakly suppresses
  19. L1:1739 weakly excites
  20. L2:1327 weakly excites
  21. L2:2894 weakly excites
  22. L1:2920 weakly suppresses
  23. L2:110 weakly suppresses
  24. L2:2558 weakly suppresses

Feeds later layers

  1. L4:809 suppresses
  2. L14:1616 weakly excites
  3. L4:3003 weakly excites
  4. L10:1683 weakly excites
  5. L4:1012 weakly suppresses
  6. L16:2287 weakly suppresses
  7. L10:2919 weakly suppresses
  8. L24:2173 weakly excites
  9. L13:404 weakly suppresses
  10. L17:1965 weakly excites
  11. L20:872 weakly suppresses
  12. L4:3039 weakly suppresses
  13. L13:905 weakly suppresses
  14. L15:608 weakly excites
  15. L18:2722 weakly suppresses
  16. L12:410 weakly excites
  17. L6:613 weakly suppresses
  18. L22:411 weakly excites
  19. L16:2879 weakly suppresses
  20. L6:1689 weakly suppresses
  21. L14:2062 weakly excites
  22. L15:1084 weakly excites
  23. L11:1007 weakly suppresses
  24. L10:1991 weakly suppresses the word 'voluntary'

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.

2 values between -2.57 and -2.481 values between -1.95 and -1.861 values between -1.77 and -1.681 values between -1.24 and -1.152 values between -1.15 and -1.062 values between -0.89 and -0.801 values between -0.80 and -0.711 values between -0.71 and -0.623 values between -0.62 and -0.532 values between -0.53 and -0.4410 values between -0.44 and -0.3673 values between -0.36 and -0.27260 values between -0.27 and -0.18645 values between -0.18 and -0.091123 values between -0.09 and -0.001064 values between -0.00 and 0.09485 values between 0.09 and 0.18164 values between 0.18 and 0.2635 values between 0.26 and 0.3514 values between 0.35 and 0.446 values between 0.44 and 0.536 values between 0.53 and 0.622 values between 0.62 and 0.712 values between 0.71 and 0.801 values between 0.80 and 0.882 values between 0.88 and 0.97offnegativepositiveoften

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. 扑
  2. 碉
  3. OUS
  4. ounding
  5. ValidationResult
  6. 党总支
  7. 党委书记
  8. GO
  9. uckle
  10. 执业

Suppresses

  1. yourselves
  2. __ ⏎
  3. [#
  4. Lamar
  5. yourself
  6. openness
  7. oneself
  8. himself
  9. �택
  10. favor