Neuron Explorer

Qwen3-0.6B

end of document after a full paragraph

A neuron in layer 12 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 responds strongly to the hidden end-of-text marker when it comes after a complete paragraph, so it seems to sense that one piece of writing has finished. Turning it up did not give a 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

    their . place. So, the students, teachers and other professionals finally had to resume their studies and jobs in government institutions like schools, colleges, courts etc.<|endoftext|>I hope you are all well and that you enjoyed some of your weekend in the

  2. #2 peak

    own how to get there in the same way they landed the jet five years before.⏎So, there are lessons here:⏎- Without leadership from the top, bottom<|endoftext|>Updated: Jun 1, 2020⏎Jaundice is

  3. #3 peak

    –6. Warsaw, 1877–82.⏎Pamiqtki Soplicy. Warsaw, 1961.<|endoftext|>Translate this page into:⏎Approach to family assessment and intervention⏎All India Institute

  4. #4 peak

    i>⏎;; Now insert one character at the beginning of the buffer: (goto-char (point-min)) ⇒ 1 (insert "Q") ⇒ nil⏎<|endoftext|>A bead is a small bauble or jewel with a hole through its middle for

  5. #5 peak

    to break down carbohydrates, proteins and fats.⏎You may have an idea, but there are many kinds of digestive enzymes that break down different kinds of food items.<|endoftext|>Chinese architecture started from ancient cave-dwellings, that are still inhabited on the

  6. #6 peak

    editing nightmare and when the sorting is finished, it will put all of the blank spaces at the top of the list, leaving the data single-spaced anyway.<|endoftext|>A hydrogen atom falling on a neutron star surface from infinity releases 200

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: X. A. So I Have. A. So A. Therefore. So A. Therefore. A. So A. Therefore. So A. Therefore. So A. Therefore. So A. Therefore. So A. Therefore. So

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

    Countess of Rosebery, died, at Dalmeny House, in Scotland, at the

  2. #2 peak

    systems.⏎What is Lyra's demon?⏎lyra's demon, pantalaimon /ˌpæntəˈlaɪmən/, is his most beloved companion, whom he calls “

  3. #3 peak

    10101100.0001000x.xxxxxxxx.xxxxxxxx SUBNET IP: 10101100.0001000x.xyyyyyyy.y

  4. #4 peak

    Holiday Crowd At The Zoo⏎- The Child Is Father Of The Man (Es⏎Past Perfect Tense Interrogative⏎Had + subject + verb 3rd from + object.⏎Past Perfect Interrogative

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. L10:174 weakly excites
  2. L9:1423 weakly excites
  3. L4:1697 weakly excites
  4. L0:1414 weakly excites
  5. L8:1567 weakly suppresses
  6. L7:2237 weakly excites
  7. L8:1290 weakly suppresses
  8. L4:542 weakly suppresses
  9. L9:428 weakly excites split word pieces in names and years
  10. L10:229 weakly excites
  11. L1:2833 weakly excites
  12. L1:2458 weakly excites
  13. L6:2296 weakly excites
  14. L10:348 weakly suppresses
  15. L5:2422 weakly suppresses
  16. L4:2470 weakly suppresses
  17. L9:1531 weakly excites
  18. L5:1693 weakly suppresses
  19. L10:2654 weakly suppresses
  20. L5:1062 weakly suppresses
  21. L3:2490 weakly excites
  22. L0:555 weakly suppresses
  23. L9:1589 weakly suppresses
  24. L7:2364 weakly suppresses

Feeds later layers

  1. L17:712 weakly excites
  2. L19:809 weakly suppresses
  3. L14:2525 weakly excites
  4. L27:1696 weakly suppresses
  5. L14:177 weakly excites
  6. L22:2753 weakly suppresses
  7. L17:32 weakly excites end of document, boilerplate footers
  8. L18:783 weakly suppresses
  9. L20:750 weakly suppresses unclear: chinese academic text, 'when' loops
  10. L21:2151 weakly suppresses
  11. L26:2780 weakly suppresses
  12. L18:1679 weakly suppresses
  13. L15:1601 weakly suppresses
  14. L21:250 weakly excites
  15. L15:2463 weakly suppresses
  16. L23:2922 weakly suppresses
  17. L24:174 weakly excites
  18. L15:448 weakly suppresses
  19. L18:119 weakly excites
  20. L14:469 weakly excites
  21. L13:1001 weakly suppresses
  22. L23:734 weakly excites
  23. L17:57 weakly excites
  24. L14:2152 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 -1.30 and -0.897 values between -0.89 and -0.481518 values between -0.48 and -0.072311 values between -0.07 and 0.3459 values between 0.34 and 0.752 values between 0.75 and 1.161 values between 1.16 and 1.571 values between 7.71 and 8.121 values between 8.12 and 8.531 values between 8.53 and 8.941 values between 8.94 and 9.351 values between 10.57 and 10.981 values between 11.80 and 12.211 values between 13.03 and 13.441 values between 14.67 and 15.08offnegativepositiveoften

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. !
  2. ax
  3. ady
  4. ax
  5. (ax
  6. asjon
  7. :
  8. ę
  9. ?
  10. [

Suppresses

  1. ��
  2. '\\'
  3. Serviço
  4. "\\"
  5. 经营模式
  6. TURE
  7. (','
  8. 嘉年
  9. .Resume
  10. hire