Neuron Explorer

Qwen3-0.6B

"to" after "the best way is"

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 lights up on the word "to" when a sentence is about to explain how to do something, such as "the best way to get started is to...". When it was turned up, the model's replies drifted into unrelated topics like maths and coding, so the steering result tells us little about the neuron.

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

    three Rs—reduce, reuse and recycle.⏎One way to reduce is to stop junk mail from being delivered to your home. The Bay Area Outreach Coalition offers

  2. #2 peak

    only recourse. To introduce a dummy load, the stamper typically adds a pressure system or coining station to balance the loads.⏎A better solution would be to address<|endoftext|>

  3. #3 peak

    . Using harsh cleansers and bath products including soap also worsen the problem by stripping away the lipids and oils that keep skin moist. A better alternative is to use a soap-free cleanser which contains ingredients that remove dirt while helping to repair

  4. #4 peak

    chest. They can have gray patches, however, which sometimes causes them to be mistaken for their cousin the gray fox. The best way to tell the difference is to check the tail: gray foxes have a black tipped tails<|endoftext|>ARC/INFO export

  5. #5 peak

    bin and don’t add too many food scraps to it at one time, it won’t smell or create indoor pest problems.⏎Another way to fertilize naturally is to mulch around your plants with organic matter such as grass clippings. The

  6. #6 peak

    create new passwords and actually remember them all.⏎Remember 100 different passwords with 1 rule⏎One way to generate unique passwords that you can remember is to choose a base password and then include part of the name

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: The number of the number of the number of the number of the number of the number of the number of the number of the number of the number of the number of the number of the number of the number of the number of the number of

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

    Test (OLST)1,2 is a simple, easy and effective method to screen for balance impairments in the older adult population.⏎You may be asking yourself, "how can standing on one leg provide you with any information about balance,

  2. #2 peak

    influences from Italian opera, and embraced the emfindsamer stil ideals popular in the mid-18thC. You can hear each of these influences in her Harpsichord Concerto in A Major,

  3. #3 peak

    and support one another⏎Encourage person to explain reasons for objection⏎Refocus his/her participation as recorder.⏎(5) The side conversationalist:⏎Enforce ground rules⏎Use pause and look technique⏎Invite the person to share views with

  4. #4 peak

    Establishing a program of incentives and penalties is not enough, alone, however. Enforceability is important—the market alone can’t make improvements permanent. Questions also remain 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. L0:2362 weakly suppresses
  2. L0:2268 weakly suppresses
  3. L0:371 weakly suppresses
  4. L0:1179 weakly suppresses
  5. L0:773 weakly excites
  6. L0:370 weakly excites
  7. L0:3054 weakly suppresses
  8. L0:2394 weakly suppresses
  9. L0:2216 weakly excites
  10. L0:768 weakly suppresses
  11. L0:2258 weakly suppresses
  12. L0:2107 weakly excites
  13. L0:581 weakly excites
  14. L0:2541 weakly suppresses
  15. L0:1038 weakly excites
  16. L0:933 weakly excites
  17. L0:248 weakly excites
  18. L0:368 weakly suppresses
  19. L0:325 weakly excites
  20. L0:2782 weakly excites
  21. L0:178 weakly suppresses
  22. L0:226 weakly excites
  23. L0:1638 weakly suppresses
  24. L0:1301 weakly suppresses

Feeds later layers

  1. L5:293 excites
  2. L20:427 weakly suppresses
  3. L26:309 weakly excites
  4. L11:1208 weakly excites
  5. L22:2094 weakly excites
  6. L27:2391 weakly suppresses
  7. L27:479 weakly suppresses
  8. L15:2363 weakly excites
  9. L7:949 weakly excites
  10. L19:2927 weakly suppresses
  11. L19:445 weakly excites
  12. L4:2259 weakly excites
  13. L14:1227 weakly excites
  14. L23:803 weakly suppresses
  15. L5:1654 weakly suppresses
  16. L6:383 weakly excites
  17. L4:562 weakly excites blank slot between a date's day and year
  18. L12:682 weakly excites
  19. L16:1759 weakly suppresses
  20. L11:992 weakly suppresses
  21. L7:1897 weakly suppresses
  22. L25:894 weakly excites
  23. L7:403 weakly excites
  24. L10:483 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.

2 values between -0.88 and -0.793 values between -0.79 and -0.701 values between -0.70 and -0.614 values between -0.61 and -0.528 values between -0.52 and -0.4315 values between -0.43 and -0.3449 values between -0.34 and -0.25145 values between -0.25 and -0.16403 values between -0.16 and -0.071326 values between -0.07 and 0.021214 values between 0.02 and 0.11546 values between 0.11 and 0.20146 values between 0.20 and 0.2922 values between 0.29 and 0.387 values between 0.38 and 0.472 values between 0.47 and 0.561 values between 0.56 and 0.654 values between 0.65 and 0.743 values between 0.74 and 0.831 values between 0.83 and 0.921 values between 1.27 and 1.362 values between 1.54 and 1.631 values between 1.90 and 1.991 values between 2.26 and 2.351 values between 2.62 and 2.71offnegativepositiveoften

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. est
  2. (OP
  3. 赤
  4. osc
  5. trí
  6. LAS
  7. He
  8. opt
  9. astically
  10. \

Suppresses

  1. ationship
  2. 架子
  3. 管局
  4. ographed
  5. mãi
  6. tered
  7. credit
  8. waukee
  9. 门
  10. Ministry