Neuron Explorer

Qwen3-0.6B

"of" after "board" or "off"

A neuron in layer 1 of Qwen3-0.6B. Its strongest snippets are very consistent: they mostly peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on the word "of" when it follows "Board" or "off", as in "Board of Education". Its strongest negative reaction is on "kind" in phrases like "one of a kind". The steering results do 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

    the Woman's Board of Missions for the Pacific Islands, located in Hawaii (WBMPI). The three mainland boards cooperated with the maleled American Board of Commissioners for Foreign Missions.(2) The women's boards chose their missionaries and

  2. #2 peak

    action programs.⏎A crucial turning point came in 1995, when the University of California’s Board of Regents made the calamitous decision to eliminate affirmative action for its then nine universities,

  3. #3 peak

    To make a series connection between batteries, use a jumper wire between the negative of the first battery and the positive of the second battery. Run your positive wire off of

  4. #4 peak

    's Millville Board of Education to 32 educators across the district's nine schools.⏎"Three hundred

  5. #5 peak

    red squirrels and porcupines.⏎Types of Flora⏎The animals of the park have plenty to forage and thrive off of with a strong botanical environment in Palo Verde National Park. Some of the plant species

  6. #6 peak

    Manitowoc Rapids. It is estimated that more than 5,000 people attended the event. Today, the marker is located off of Broadway Street.⏎In the days before settlement, our area was named Manitowoc

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:. of credit, of of financial with2ere2 with 2 of2Ins of2 financial analysisahas... of2 financials consolidated and financialn. financialn$ and x2

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

    for more sophisticated future missions.⏎SLIM: A Historic Lunar Mission⏎Taking lunar exploration to new scales, JAXA’s SLIM mission is one of a kind in the world of space pioneering. The mission is designed to control the lander

  2. #2 peak

    When your cells can’t get glucose (a kind of sugar) to burn for energy, they burn fat instead. This leaves behind

  3. #3 peak

    the steel underneath. However, hot-dip can be done on a much larger scale and provides extremely long-lasting protection and<|endoftext|>Placenta forceps are a kind of gynecological instruments. These forceps are used after the mother has given

  4. #4 peak

    the condition suitable for consumption, to make them as tasty as possible, and unleash their maximum potential.⏎The process takes place in a roaster, which is a kind of oven that transfers heat to the beans by convection, conduction, and

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. L0:1301 strongly suppresses
  2. L0:1224 strongly suppresses
  3. L0:178 strongly suppresses
  4. L0:1564 weakly suppresses
  5. L0:191 weakly suppresses
  6. L0:1638 weakly excites
  7. L0:1 weakly suppresses
  8. L0:2003 weakly excites
  9. L0:2883 weakly excites
  10. L0:2610 weakly suppresses
  11. L0:2340 weakly suppresses
  12. L0:157 weakly suppresses
  13. L0:2701 weakly suppresses
  14. L0:2566 weakly suppresses
  15. L0:1075 weakly excites
  16. L0:1116 weakly excites
  17. L0:109 weakly excites
  18. L0:1131 weakly excites
  19. L0:2784 weakly excites
  20. L0:1951 weakly excites
  21. L0:391 weakly suppresses
  22. L0:455 weakly excites
  23. L0:2641 weakly suppresses
  24. L0:2492 weakly suppresses

Feeds later layers

  1. L2:1628 suppresses
  2. L3:59 suppresses
  3. L18:46 excites
  4. L12:1945 suppresses
  5. L16:213 excites
  6. L2:2730 suppresses
  7. L16:182 excites
  8. L3:516 weakly suppresses
  9. L3:102 weakly suppresses
  10. L5:3001 weakly suppresses
  11. L12:1767 weakly suppresses
  12. L12:1472 weakly suppresses
  13. L7:1 weakly excites
  14. L3:2911 weakly suppresses
  15. L5:46 weakly suppresses
  16. L18:57 weakly excites
  17. L3:115 weakly suppresses
  18. L23:1646 weakly suppresses
  19. L21:1766 weakly suppresses 'in' in 'bear in mind' phrases
  20. L2:3015 weakly suppresses
  21. L2:1050 weakly suppresses
  22. L3:2042 weakly suppresses
  23. L7:122 weakly suppresses
  24. L3:1180 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 -0.45 and -0.423 values between -0.31 and -0.283 values between -0.28 and -0.259 values between -0.25 and -0.2116 values between -0.21 and -0.1815 values between -0.18 and -0.1429 values between -0.14 and -0.1177 values between -0.11 and -0.08167 values between -0.08 and -0.04525 values between -0.04 and -0.011041 values between -0.01 and 0.03702 values between 0.03 and 0.06407 values between 0.06 and 0.09269 values between 0.09 and 0.13196 values between 0.13 and 0.16114 values between 0.16 and 0.2089 values between 0.20 and 0.2368 values between 0.23 and 0.2642 values between 0.26 and 0.3033 values between 0.30 and 0.3318 values between 0.33 and 0.3720 values between 0.37 and 0.4013 values between 0.40 and 0.4311 values between 0.43 and 0.477 values between 0.47 and 0.504 values between 0.50 and 0.545 values between 0.54 and 0.578 values between 0.57 and 0.605 values between 0.60 and 0.642 values between 0.64 and 0.671 values between 0.67 and 0.702 values between 0.70 and 0.741 values between 0.74 and 0.772 values between 0.77 and 0.811 values between 0.84 and 0.872 values between 0.87 and 0.91offnegativepositiveoften

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. .Free
  2. ../../../
  3. v
  4. 个工作
  5. たり
  6. urther
  7. preced
  8. erd
  9. <<
  10. 趯

Suppresses

  1. Builders
  2. despre
  3. elts
  4. pellets
  5. 槿
  6. Phelps
  7. urtles
  8. Phones
  9. $filter
  10. -Sah