Neuron Explorer

Qwen3-0.6B

unclear: document ends and math commas

A neuron in layer 23 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 is hard to pin down. It fires most strongly where a web page's text seems to cut off and the next document begins, and it goes strongly negative on commas in maths problems. Turning it down made the model repeat the word 'find' and turning it up gave ordinary sentences, so the pattern is not clear.

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

    31 minutes outside the lab.⏎After what was then a record 84 days in space, Carr, Pogue and Gibson bid farewell to Skylab and<|endoftext|>By

  2. #2 peak

    joined to form a⏎quadrilateral. What do you notice about this quadrilateral as the⏎dynamic image changes? Can you prove your conjecture?⏎It is impossible to trisect an angle using

  3. #3 peak

    decision.⏎Each branch must depend on the other in order to keep things running smoothly. Many times, this presents challenges along the way. It also means that it<|endoftext|>Writers Building Kolkata⏎Writer’s Building, Kolkata⏎Going Back to the History

  4. #4 peak

    so that cows can access shade and water whenever they want it. But that is easier said than done. Over the next ten years, we will be expected to<|endoftext|>

  5. #5 peak

    . Their music and lyrics are synonymous with the 60’s. From “The Sound of Silence” to “Bridge Over Troubled Water”, their songs act<|endoftext|>Read Music Notes⏎How to Read All the Music Notes -- With Ease⏎So

  6. #6 peak

    resistant” may be a special treat to hungry deer, as they munch on tender soft new growth. Like many other animals, deer have food they avoid, and<|endoftext|>Seven miles to the south west of Sheffield there's a piece of wild moor

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: find find find find find find find find find find find find find find find find find find find find find find find find find find find find find find find find find find find find find find find find find find find find find find find find

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

    all variables and given/known data If an athlete jumped 2 feet high and left the ground at an angle of 20 degrees with respect to the horizontal, how fast was the athlete going in the forward (positive horizontal) and upward (

  2. #2 peak

    centripetal acceleration at point B?⏎Fig.1 above refer to a point moving along a circular path. If BD is the diameter of the circle of rotation, then the velocities at A and B are

  3. #3 peak

    .⏎A straight line is drawn between the end points of the arc would be a⏎chord of the circle.⏎If the arc length is exactly half the circle, this called a semicircular arc. See Semicircle definition.⏎Naming and

  4. #4 peak

    .⏎An ideal monatomic gas initially at 298 K and 1 atm pressure is⏎expanded from a volume of 1 liter to 2 liters. Consider separately⏎isobaric, isothermal and adiabatic expansions to

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. L17:2296 suppresses
  2. L18:1586 suppresses
  3. L15:1416 suppresses
  4. L21:728 suppresses
  5. L19:877 excites
  6. L20:2335 excites comma ending an 'if' clause
  7. L3:66 suppresses
  8. L7:61 excites
  9. L18:2010 excites
  10. L3:52 suppresses
  11. L9:128 suppresses
  12. L19:2676 excites
  13. L20:810 suppresses
  14. L15:2327 suppresses
  15. L11:1483 suppresses
  16. L4:150 weakly suppresses
  17. L13:2989 weakly excites nouns for who or what a text is about
  18. L6:431 weakly suppresses
  19. L9:156 weakly excites
  20. L8:584 weakly excites
  21. L8:49 weakly suppresses
  22. L11:1843 weakly excites
  23. L20:1086 weakly suppresses
  24. L12:1069 weakly excites

Feeds later layers

  1. L24:1302 suppresses
  2. L27:1284 suppresses
  3. L26:1282 suppresses
  4. L27:247 suppresses
  5. L25:361 suppresses
  6. L26:623 suppresses
  7. L25:364 weakly suppresses
  8. L26:1422 weakly suppresses
  9. L26:675 weakly suppresses
  10. L24:837 weakly suppresses
  11. L26:2988 weakly suppresses
  12. L24:290 weakly suppresses 'to' in set phrases like 'to come'
  13. L25:1335 weakly suppresses
  14. L26:1104 weakly suppresses
  15. L24:2640 weakly suppresses
  16. L26:939 weakly suppresses
  17. L26:671 weakly suppresses
  18. L27:900 weakly suppresses
  19. L27:645 weakly suppresses
  20. L26:296 weakly suppresses
  21. L26:2821 weakly suppresses
  22. L24:2579 weakly suppresses
  23. L25:660 weakly suppresses
  24. L24:2716 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 -13.40 and -12.851 values between -11.21 and -10.671 values between -9.03 and -8.481 values between -8.48 and -7.931 values between -7.39 and -6.841 values between -6.84 and -6.291 values between -6.29 and -5.753 values between -5.75 and -5.201 values between -5.20 and -4.662 values between -4.66 and -4.115 values between -4.11 and -3.561 values between -3.56 and -3.026 values between -3.02 and -2.476 values between -2.47 and -1.9215 values between -1.92 and -1.3845 values between -1.38 and -0.83426 values between -0.83 and -0.282581 values between -0.28 and 0.26741 values between 0.26 and 0.8139 values between 0.81 and 1.369 values between 1.36 and 1.904 values between 1.90 and 2.456 values between 2.45 and 3.002 values between 3.00 and 3.541 values between 3.54 and 4.092 values between 4.09 and 4.641 values between 4.64 and 5.181 values between 5.18 and 5.732 values between 6.28 and 6.821 values between 6.82 and 7.371 values between 7.91 and 8.46offnegativepositiveoften

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. yst
  2. ivalence
  3. 本报讯
  4. aves
  5. ibox
  6. ipped
  7. 批
  8. offee
  9. 깨
  10. ierten

Suppresses

  1. find
  2. find
  3. 求
  4. Find
  5. Find
  6. .find
  7. _find
  8. (find
  9. -find
  10. determine