Neuron Explorer

Qwen3-0.6B

intensifiers like 'largely' and 'backbone'

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

This neuron weakly fires on words that describe degree or importance, such as 'largely', 'widely', 'both', 'backbone', and 'cornerstone', usually in the middle of a sentence. It is weak compared with the strongest neurons in this layer. Steering produced loops and unrelated text, so it does not show a clear meaning.

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

    precision and creativity, as developers navigate through various programming languages and frameworks to create powerful and intuitive applications.⏎The Significance of Coding in App Development⏎Coding is the backbone of app development, allowing developers

  2. #2 peak

    -spectrum beta-lactamase with additional resistance to carbenicillin, amoxicillin, and increasingly to ceftazidime. The bacteria remain largely susceptible to aminoglycosides and cephalosporins. V

  3. #3 peak

    ched structure from the Gothic church is characteristic of this specific style of architecture. Gothic cathedrals were typically designed with an individual wall, which has been both decorated or plain. The interior had been often decorated using tapestries and pictures

  4. #4 peak

    The lower number may<|endoftext|>Combining optimization with analytics is a frequent topic now. Through data analytics and optimizations are specifically hot topics at the moment, they’ve been widely used in marketing analytics and research for over 30 years.

  5. #5 peak

    is larger than, for example, the accommodation and food services sector; the utilities sector; and the arts, entertainment and recreation sector.⏎Agriculture is a cornerstone industry that supports other economic activity, provides an essential product and is poised for long

  6. #6 peak

    . Election day was largely peaceful and approximately 81% of voters cast ballots, one of the highest

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 a new question from a user, and I have a new question from a user. So, I have a new question From a new question, a new question from a new user: user: a new question: a new

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

    production. But how is protein production controlled?⏎Researchers of the Max Delbruck Center for Molecular Medicine (MDC) Berlin-Buch of the Helmholtz Association, Germany, have now comprehensively quantified gene expression (the activation of

  2. #2 peak

    theme that remains constant throughout the play is appearance versus reality.⏎Four of the main characters that this theme is best seen through are Polonius, Rosencrantz, Guildenst

  3. #3 peak

    level of relative sea-level rise, when they should be building projects that can function for several generations as sea level rises twice as high, if not higher.”⏎Garret Graves, head of the state Coastal Planning and Protection Authority, did not respond to

  4. #4 peak

    can reveal more of its flaws.⏎The baguette cut of diamond is usually used for side stones for other cuts of diamonds, 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. L1:2679 weakly excites
  2. L0:37 weakly excites
  3. L1:1279 weakly suppresses
  4. L1:2235 weakly suppresses
  5. L2:49 weakly suppresses
  6. L1:1786 weakly suppresses
  7. L2:2563 weakly suppresses
  8. L1:1875 weakly excites
  9. L0:571 weakly excites
  10. L2:2015 weakly suppresses
  11. L0:1415 weakly excites
  12. L2:2643 weakly suppresses
  13. L1:2562 weakly excites
  14. L0:2447 weakly excites
  15. L1:685 weakly suppresses
  16. L2:930 weakly excites
  17. L0:1347 weakly excites
  18. L0:2518 weakly suppresses
  19. L1:1219 weakly excites
  20. L2:2727 weakly excites
  21. L2:957 weakly suppresses
  22. L1:2557 weakly excites
  23. L2:2713 weakly suppresses
  24. L2:154 weakly excites

Feeds later layers

  1. L14:401 weakly excites
  2. L13:1599 weakly suppresses
  3. L20:1071 weakly excites
  4. L12:1820 weakly suppresses
  5. L5:398 weakly suppresses
  6. L12:18 weakly excites
  7. L14:399 weakly excites
  8. L16:1102 weakly suppresses
  9. L22:477 weakly suppresses
  10. L16:1958 weakly excites
  11. L6:2767 weakly excites
  12. L11:426 weakly excites
  13. L12:1834 weakly suppresses
  14. L7:105 weakly excites
  15. L17:1861 weakly suppresses
  16. L18:1053 weakly suppresses
  17. L8:2699 weakly excites
  18. L8:400 weakly suppresses
  19. L14:1594 weakly suppresses
  20. L7:894 weakly suppresses
  21. L26:592 weakly suppresses
  22. L8:616 weakly excites
  23. L10:3033 weakly excites
  24. L24:408 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 -0.71 and -0.661 values between -0.62 and -0.572 values between -0.53 and -0.484 values between -0.48 and -0.447 values between -0.44 and -0.4010 values between -0.40 and -0.3519 values between -0.35 and -0.3126 values between -0.31 and -0.2680 values between -0.26 and -0.22149 values between -0.22 and -0.18229 values between -0.18 and -0.13394 values between -0.13 and -0.09461 values between -0.09 and -0.04817 values between -0.04 and 0.00682 values between 0.00 and 0.05406 values between 0.05 and 0.09244 values between 0.09 and 0.14174 values between 0.14 and 0.1874 values between 0.18 and 0.2253 values between 0.22 and 0.2732 values between 0.27 and 0.3119 values between 0.31 and 0.367 values between 0.36 and 0.403 values between 0.40 and 0.441 values between 0.44 and 0.493 values between 0.49 and 0.532 values between 0.53 and 0.582 values between 0.62 and 0.671 values between 0.71 and 0.751 values between 0.89 and 0.932 values between 0.93 and 0.982 values between 1.02 and 1.06offnegativepositiveoften

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. 鸥
  2. Od
  3. square
  4. vu
  5. uada
  6. imulator
  7. Od
  8. ollower
  9. 数据库
  10. lector

Suppresses

  1. __;
  2. posé
  3. privation
  4. alon
  5. dotyczące
  6. änge
  7. �回
  8. сон
  9. APTER
  10. ntax