Neuron Explorer

Qwen3-0.6B

opening parenthesis before an acronym

A neuron in layer 26 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on the opening bracket that comes right before a short acronym, such as the '(' in 'Resting Metabolic Rate (RMR)'. Turned down, the model produced garbled repeated words. Turned up, it wrote text that did not clearly match.

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

    during the oxidation of food and carbon dioxide (CO ) is produced. 2⏎), in proportion to the heat production.⏎Measuring Resting Metabolic Rate (RMR) is fast and simple. It includes a short, non-invasive

  2. #2 peak

    made in laboratories around the world. And they are among the goals of more than 100 horizon-stretching research projects housed in the Advanced Measurement Laboratory (AML) Complex at the National Institute of Standards and Technology (NIST).⏎Completed

  3. #3 peak

    of progressive science education centres within the Africa-Mediterranean-European (AME) area.<|endoftext|>THE NEW FOUNDE LAND⏎Volume 18

  4. #4 peak

    , and gave glycerin-supplemented poultry feed to broilers that were 7 to 45 days old.⏎Dozier evaluated the apparent metabolizable energy (AME) intake—a standard measure of energy—for the study group and for a control

  5. #5 peak

    vibrations (<100 Hz) are naturally generated by skeletal muscles during voluntary contractions. Recording of these vibrations at the muscle surface are called surface mechanomyograms (S-MMGs). In this study, S-MMGs were recorded over a

  6. #6 peak

    . https://www.academia.edu/39903340/Displaying_an_Icon_The_Mosaic_Icon_of_Saint_Demetrios_at_Sassoferrato_and_its_Frame

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: remmorum remmorum rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem rem

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

    After the Romans London was conquered by the Angles and the Saxons who brought their language with them (Old English)⏎In 1066 William the Conqueror conquered London with his Norman (though French speaking) army. Not only

  2. #2 peak

    quest, documented history starts around the end of the 13 th century, when the hall was passed to the Earl of Chester, a nephew of William the Conqueror.⏎Since the year

  3. #3 peak

    "?⏎Yes, curling has been compared to chess by fans and detractors alike. Fans emphasize the strategy involved. Detractors claim that both sports

  4. #4 peak

    hands, (Poland having been invaded by Germany), he has hidden it.<|endoftext|>BURKE TEACHERS STUDY, PREPARE FOR NEW CURRICULUM⏎Burke County schools are preparing for a curriculum change coming in August 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. L25:1061 strongly suppresses
  2. L22:1810 excites
  3. L25:925 excites
  4. L25:2560 suppresses
  5. L24:1201 excites
  6. L25:237 excites
  7. L24:2795 excites
  8. L21:133 suppresses
  9. L23:1793 weakly excites
  10. L19:334 weakly excites
  11. L19:1581 weakly suppresses
  12. L23:1334 weakly suppresses
  13. L24:1038 weakly suppresses
  14. L5:1397 weakly suppresses
  15. L0:2560 weakly excites
  16. L12:1357 weakly excites
  17. L10:2114 weakly suppresses
  18. L6:940 weakly excites
  19. L12:3001 weakly excites
  20. L17:1059 weakly suppresses
  21. L4:2018 weakly excites
  22. L25:2378 weakly suppresses
  23. L23:188 weakly excites
  24. L7:1368 weakly excites

Feeds later layers

  1. L27:422 suppresses
  2. L27:464 suppresses
  3. L27:1012 suppresses
  4. L27:941 suppresses
  5. L27:211 excites
  6. L27:790 weakly suppresses
  7. L27:3027 weakly suppresses
  8. L27:2995 weakly suppresses
  9. L27:2288 weakly suppresses
  10. L27:2608 weakly suppresses
  11. L27:1515 weakly suppresses
  12. L27:720 weakly suppresses
  13. L27:1933 weakly suppresses
  14. L27:89 weakly excites
  15. L27:87 weakly suppresses
  16. L27:2771 weakly suppresses
  17. L27:2796 weakly suppresses
  18. L27:2178 weakly suppresses
  19. L27:1359 weakly suppresses
  20. L27:590 weakly excites
  21. L27:178 weakly excites
  22. L27:2596 weakly suppresses
  23. L27:1426 weakly suppresses
  24. L27:991 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 -20.55 and -19.461 values between -15.11 and -14.021 values between -12.94 and -11.853 values between -10.76 and -9.682 values between -9.68 and -8.592 values between -8.59 and -7.505 values between -7.50 and -6.413 values between -6.41 and -5.339 values between -5.33 and -4.2413 values between -4.24 and -3.1525 values between -3.15 and -2.0761 values between -2.07 and -0.982770 values between -0.98 and 0.11891 values between 0.11 and 1.2076 values between 1.20 and 2.2820 values between 2.28 and 3.3715 values between 3.37 and 4.464 values between 4.46 and 5.542 values between 6.63 and 7.722 values between 7.72 and 8.811 values between 8.81 and 9.891 values between 21.85 and 22.94offnegativepositiveoften

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. <!--[
  3. <!--[
  4. ưỡ
  5. ▩
  6. 只会
  7. hasNext
  8. canActivate
  9. 뜨
  10. viewDidLoad

Suppresses

  1. em
  2. Emm
  3. rem
  4. EM
  5. rem
  6. Em
  7. EM
  8. Em
  9. um
  10. (em