Neuron Explorer

Qwen3-0.6B

"This" starting a labelled sentence

A neuron in layer 0 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 "This" when it begins a new sentence right after a label or colon, like "Achilles Tendon: This tendon..." Its negative side is weak and shows up on colons at the start of a text. Steering did not show a clear effect tied to this pattern.

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

    possibilities:⏎- This may mean that you are one of the 15-20 percent of

  2. #2 peak

    Initiative Prevention Curriculum⏎Target Audience: Middle school-aged youth⏎Administration: This curriculum can be used by any agency that works with youth and that has a mission

  3. #3 peak

    .⏎Nephrotic Syndrome: This condition causes the kidney to secrete extreme proteins from the body by urine. Our

  4. #4 peak

    to deal with all that other stuff, you can plant these trees.⏎Chinese Pistache (Pistache chinesis): This is an ornamental species, meaning that it will not produce pistachio nuts

  5. #5 peak

    it could be done as...⏎Demonstration: This simple experiment gives a reasonable value for the density of air.⏎Demonstration: This

  6. #6 peak

    - Actual Units Sold: This is the actual sales performance during the period.⏎- Budgeted Units Sold: The

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 am not the one who is the one I have to be the one who is the one who is the one who is the one who is the one who is the one who is the one who is the one who is the one who

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

    : Biological Safety Cabinets In Action⏎A Biological Safety Cabinets is a ventilated container that

  2. #2 peak

    : Pennsylvania Native Plant Society Demonstration Garden, Monarch Waystation<|endoftext|>Entomological

  3. #3 peak

    : or ART, can refer to medication taken by people with HIV to reduce the amount

  4. #4 peak

    : Mining World⏎Population: 300 (Custodial staff)⏎Ge

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

None: this is the first layer.

Feeds later layers

  1. L12:1062 suppresses
  2. L6:1121 weakly suppresses 'while' joining two clauses after a comma
  3. L1:713 weakly excites "a" after a name, introducing a title
  4. L1:1446 weakly suppresses
  5. L3:595 weakly suppresses
  6. L7:790 weakly suppresses
  7. L11:2691 weakly suppresses
  8. L5:1617 weakly suppresses
  9. L22:178 weakly excites
  10. L18:1281 weakly suppresses
  11. L6:969 weakly suppresses
  12. L9:1193 weakly suppresses
  13. L2:1466 weakly suppresses
  14. L5:1333 weakly excites
  15. L11:1153 weakly suppresses
  16. L6:662 weakly suppresses
  17. L8:1519 weakly suppresses
  18. L19:1492 weakly suppresses
  19. L21:265 weakly excites
  20. L15:1746 weakly suppresses
  21. L7:497 weakly suppresses
  22. L14:422 weakly suppresses periods after initials and abbreviations
  23. L16:1115 weakly excites
  24. L12:2929 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 -0.55 and -0.502 values between -0.33 and -0.287 values between -0.28 and -0.2212 values between -0.22 and -0.1756 values between -0.17 and -0.11192 values between -0.11 and -0.061312 values between -0.06 and 0.001769 values between 0.00 and 0.06447 values between 0.06 and 0.1168 values between 0.11 and 0.1720 values between 0.17 and 0.2212 values between 0.22 and 0.282 values between 0.28 and 0.331 values between 0.33 and 0.391 values between 0.39 and 0.441 values between 0.44 and 0.501 values between 0.50 and 0.551 values between 0.55 and 0.611 values between 0.61 and 0.661 values between 0.83 and 0.891 values between 1.60 and 1.66offnegativepositiveoften

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. fragment
  2. 몰
  3. eden
  4. piece
  5. eken
  6. yan
  7. ens
  8. je
  9. 中原
  10. arten

Suppresses

  1. cabinets
  2. SPATH
  3. figures
  4. bè
  5. quila
  6. NCY
  7. posables
  8. cases
  9. uracy
  10. 这条路