Neuron Explorer

Qwen3-0.6B

colon that opens a list

A neuron in layer 18 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 colon at the end of a sentence that introduces a list or set of steps, such as 'Try these tips:'. Its negative side fires inside odd word fragments, which is hard to make sense of. Steering showed no clear theme.

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

    assessors). Here, some of the most important considerations to be taken into account when writing a technical report are highlighted. This section will briefly address the following points:⏎What is a technical report?⏎In simple terms, a report is a document that

  2. #2 peak

    . If you cheat, your transaction does not get forwarded.⏎I hope that explains the concept of how Bitcoin works in regards to your question. Just to be clear:⏎"Does each bitcoin has a unique identifier,

  3. #3 peak

    the origins of health and also the etiology, treatment and prevention of many of the chronic diseases currently affecting the Canadian population. Specific CEEHRC objectives include:⏎- Establishment of a national epigenetics platform that will serve as an essential resource

  4. #4 peak

    argument worked; sometimes it didn’t.⏎Current arguments against the legality of mask mandates⏎Here are examples of anti-maskers’ arguments for why mask mandates are illegal:⏎- Costco Karen, as she is called, staged a sit-in at a Hills

  5. #5 peak

    . Take a look below at some of the most common questions that we hear from educators and the solution to solve them:⏎"Where do I start?"⏎You are an educator who either purchased your own classroom

  6. #6 peak

    are severe, a condition known as hyperemesis gravidarum may be diagnosed.⏎Eating and Drinking⏎Changing what you eat may help. Try these tips:⏎- Eat a lot of protein and carbohydrates. Try peanut butter on apple slices or

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: If the number of possible values of the function's value on the integer integer integer integer integer integer integer integer integer integer integer integer integer integer integer integer integer integer, and the number of the function's values on the more complicated integer integer integer,

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

    ities reproduced in⏎Rice: Amber the Golden Gem of the Ages, the Polish Academy of Sciences:⏎Amber in Nature, a German⏎publication: Lausitzer Bernstein, and Grimaldi:⏎Amber Window to the Past.⏎Mark

  2. #2 peak

    Dussehra-Vijaya Dashami-Dasara Festival Story, Why Do We

  3. #3 peak

    maximum of 100 mg/day. The American Congress of Obstetricians and Gynecologists recommends that first-line treatment of nausea and<|endoftext|>PITTSBURGH — Do you have a daily indulgence in your diet? Even in moderation

  4. #4 peak

    effective and so much cheaper than removal and replacement that I can’t get a single elected official to weigh in on the side of removing healthy trees because we<|endoftext|>Tourette syndrome, or Tourette’s, can be a difficult condition to live with

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. L5:1984 strongly suppresses
  2. L16:134 strongly suppresses
  3. L1:14 strongly suppresses
  4. L2:2154 strongly suppresses
  5. L16:84 excites
  6. L0:2648 suppresses
  7. L2:1774 suppresses
  8. L17:994 suppresses
  9. L16:143 excites
  10. L1:1672 excites
  11. L16:177 suppresses zero digit inside dates and codes
  12. L15:115 excites
  13. L17:886 excites
  14. L15:89 excites
  15. L16:503 excites
  16. L0:1633 suppresses
  17. L16:126 suppresses
  18. L16:14 excites
  19. L17:542 excites
  20. L0:1380 suppresses
  21. L15:78 excites
  22. L16:227 suppresses
  23. L0:2662 excites
  24. L15:96 weakly suppresses

Feeds later layers

  1. L19:196 strongly suppresses
  2. L22:1550 strongly suppresses
  3. L20:2122 suppresses
  4. L20:2652 suppresses
  5. L27:2712 suppresses
  6. L22:2908 suppresses
  7. L27:2777 suppresses
  8. L19:1083 excites
  9. L19:1241 excites
  10. L19:2478 excites
  11. L19:974 excites
  12. L19:1370 excites
  13. L21:2828 suppresses
  14. L19:798 excites
  15. L23:2337 suppresses
  16. L26:1855 suppresses
  17. L20:1085 weakly excites
  18. L23:2124 weakly suppresses
  19. L26:2776 weakly suppresses
  20. L20:1475 weakly excites
  21. L20:1308 weakly excites the digit 3 inside years
  22. L22:2429 weakly suppresses
  23. L26:1866 weakly excites
  24. L26:1519 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 -24.72 and -24.062 values between -18.09 and -17.421 values between -16.76 and -16.101 values between -16.10 and -15.432 values between -14.11 and -13.442 values between -13.44 and -12.781 values between -12.78 and -12.123 values between -10.79 and -10.131 values between -10.13 and -9.462 values between -9.46 and -8.805 values between -8.80 and -8.141 values between -8.14 and -7.473 values between -7.47 and -6.815 values between -6.81 and -6.156 values between -6.15 and -5.487 values between -5.48 and -4.829 values between -4.82 and -4.1611 values between -4.16 and -3.4922 values between -3.49 and -2.8323 values between -2.83 and -2.1739 values between -2.17 and -1.5081 values between -1.50 and -0.84194 values between -0.84 and -0.182345 values between -0.18 and 0.491086 values between 0.49 and 1.1555 values between 1.15 and 1.81offnegativepositiveoften

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. .nz
  2. pier
  3. →⏎→⏎
  4. bone
  5. ijken
  6. 缝
  7. thác
  8. PORT
  9. 度
  10. いけ

Suppresses

  1. flen
  2. tone
  3. vier
  4. wap
  5. лу
  6. UF
  7. 师
  8. 准备好
  9. 师范
  10. ionario