Neuron Explorer

Qwen3-0.6B

unclear: mid-sentence commas, 'most' before a verb

A neuron in layer 16 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

The neuron is hard to pin down. Its mildest highs are commas in the middle of sentences and in reference lists, but its strongest reactions are negative, on the word 'most' in phrases like 'commonly affects' or 'is the most used'. Turning it up or down gave rambling text with no 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

    rhinophores are not overly large. The whitish-blue patches are distributed much as in Marcus's drawing of E. canguzua.⏎• Marcus, Er. (1955) Opisthobranchia from

  2. #2 peak

    othermia, or hemodilution related to fluid or blood administration; however, an acute coagulopathy can also occur in severely injured patients independent of, or in addition to, these factors [2,3]. Several terms are used

  3. #3 peak

    — it turned out to be the coup de grace for the team.⏎Back in 2009, at the IBM Research lab in Zurich scientists imaged, for the first time, the chemical structure of an individual molecule with unprecedented resolution,

  4. #4 peak

    . This functionality is provided solely for your convenience and is in no way intended to replace human translation. Neither BioOne nor the owners and publishers of the content make, and they explicitly disclaim, any express or implied representations or warranties of any kind

  5. #5 peak

    Cars, trucks and motorcycles produce three times the pollution of buses.⏎These comparisons assume the vehicles are not filled to capacity on average. Trains come out even better in comparison to planes when both are full. (Note that Amtrak offers discounts on

  6. #6 peak

    , 2009), it is likely that diarrhea rates will increase in this region although future research will be needed to quantify this relationship.⏎Marengo, J. A., Jones<|endoftext|>(Student produced study guide from Foss, Foss

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 problem of the problem of the problem of the problem of the problem... I feel like I feel like I feel like... maybe a possible response to that, I think I feels like I feels like... I feels like... I feels like

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

    in the affected joint. This condition most commonly affects the joints in hips, knees, hands, and spine. Rarely

  2. #2 peak

    can cause severe turmoil in a person's physical, psychological, and social health. Eating disorders most commonly affect women and occur in adolescence and young adulthood, though they can occur in

  3. #3 peak

    the most widely used server-side scripting language⏎At time of this writing, PHP is the most used server-side language (78.6%) for the websites (Source:

  4. #4 peak

    6. There is a record of a David de Boivil witnessing a charter as early as 1164.<|endoftext|>For the questions below, use the most efficient Excel formula or procedure to find your answer. Calculations by hand (or

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. L2:1563 strongly excites
  2. L10:311 strongly suppresses
  3. L0:73 strongly excites
  4. L7:291 strongly excites
  5. L4:393 strongly suppresses
  6. L1:35 suppresses
  7. L7:922 suppresses
  8. L0:844 excites
  9. L2:316 excites
  10. L7:2731 suppresses
  11. L11:1383 weakly excites
  12. L4:1039 weakly suppresses 'a' after 'play' or 'plays'
  13. L1:16 weakly excites
  14. L0:1127 weakly excites
  15. L10:273 weakly excites
  16. L4:1591 weakly suppresses
  17. L4:717 weakly suppresses
  18. L7:674 weakly suppresses
  19. L1:2709 weakly excites
  20. L7:287 weakly suppresses
  21. L12:1719 weakly suppresses
  22. L5:2149 weakly excites
  23. L14:246 weakly excites
  24. L14:1387 weakly suppresses

Feeds later layers

  1. L17:300 strongly excites
  2. L17:420 strongly excites
  3. L18:198 excites
  4. L19:183 excites
  5. L18:105 excites
  6. L18:365 excites
  7. L17:399 excites
  8. L20:234 suppresses
  9. L18:474 excites
  10. L18:103 excites
  11. L21:252 excites
  12. L18:137 excites 'the' after 'to' in fixed phrases
  13. L19:683 excites
  14. L20:200 excites
  15. L19:593 excites
  16. L17:594 excites
  17. L18:133 suppresses
  18. L22:473 excites
  19. L20:247 excites
  20. L18:553 suppresses
  21. L21:505 suppresses
  22. L23:2397 suppresses
  23. L17:378 excites
  24. L18:420 excites

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 -15.07 and -14.641 values between -14.64 and -14.222 values between -12.51 and -12.092 values between -11.66 and -11.241 values between -11.24 and -10.811 values between -10.81 and -10.382 values between -9.53 and -9.102 values between -8.25 and -7.831 values between -7.83 and -7.402 values between -7.40 and -6.971 values between -6.12 and -5.703 values between -5.70 and -5.272 values between -5.27 and -4.845 values between -4.84 and -4.422 values between -4.42 and -3.993 values between -3.99 and -3.566 values between -3.14 and -2.719 values between -2.71 and -2.2916 values between -2.29 and -1.8623 values between -1.86 and -1.4346 values between -1.43 and -1.0168 values between -1.01 and -0.58286 values between -0.58 and -0.162384 values between -0.16 and 0.27974 values between 0.27 and 0.7061 values between 0.70 and 1.122 values between 1.12 and 1.552 values between 1.55 and 1.98offnegativepositiveoften

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. ughter
  2. ��
  3. ycles
  4. ince
  5. crit
  6. empt
  7. mic
  8. angan
  9. ivate
  10. ren

Suppresses

  1. floors
  2. 天花
  3. TableRow
  4. /ubuntu
  5. probe
  6. STDMETHOD
  7. osas
  8. acher
  9. DataTask
  10. 数据库