Neuron Explorer

Qwen3-0.6B

middle pieces of split medical and foreign words

A neuron in layer 11 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron tends to fire on the middle or end piece of a long word that the model split into parts, like 'ic' in 'diverticulitis' or 'os' in 'colonoscopy'. Many of its strongest examples are medical terms, though it also fires on some place and people names. The negative side is weak and not clearly about anything. Steering did not give a 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

    , decrease the cramps and pain associated with irritable bowel syndrome (or so-called spastic colitis), prevent the development of diverticulosis and diverticulitis

  2. #2 peak

    due to endometriosis or another cause. Occasionally, the gynaecologist may refer the woman to a specialist for a colonoscopy. A colonoscopy is an examination of the inside of the gastro-intestinal tract with a telescope

  3. #3 peak

    Archer, Jeremy. "Sir G. C. Whitlock (During the 1834 Coorg Campaign)". www.archerfamily.org.uk. Archer Family. Archived from the original on 13 November 2013.

  4. #4 peak

    , a U-shaped space built into the fixture (in the case of toilets<|endoftext|>Medical-Surgical Nursing at a Glance⏎Medical-Surgical Nursing at a Glance is the

  5. #5 peak

    earlier than previously thought.⏎- Cave paintings in Spain are the oldest in the world and are sufficiently ancient to be the creations of Neandertals.⏎- Neandertals hunted birds for their fashionable feathers for thousands of years and may have exploited

  6. #6 peak

    ” (Ville d'art et d'histoire). It is also one of the “finest detours in France” network (Plus beaux détours de France). Stretching out along the banks of the River Ornain, it

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: "I have been thinking about lately: "I have been thinking about lately: "I have been thinking about lately: "I have been thinking about lately: "I have been thinking about lately: "I have been thinking about lately:

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

    10th - 12th English Language Arts⏎Lesson Plan: The State of the Union⏎Learners explore the state of the union. In this U.S. government lesson, students analyze the 1998 State of

  2. #2 peak

    -defined rights in their own capacities, as quoted in the Holy Quran:⏎“And they (women) have rights (over their husbands) similar (to those of their husbands) over them to what is reasonable, but men have a degree (of responsibility

  3. #3 peak

    According to the Center for Mindful Eating, mindful eating is all about incorporating a little bit of meditation while you eat, to help you appreciate your food as you eat it.⏎Here’s how mindful eating can improve your overall health⏎Mindful Eating

  4. #4 peak

    that are pulled through the water by the vessels. The trawls are midwater- pulled through the water column-, and bottom- dragged along the seabed. The bottom trawling

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. L9:1110 excites
  2. L10:605 suppresses
  3. L6:267 suppresses
  4. L8:78 suppresses
  5. L2:51 excites
  6. L6:135 excites
  7. L10:1062 suppresses
  8. L6:180 excites
  9. L8:891 excites
  10. L5:299 suppresses
  11. L5:711 excites
  12. L8:124 suppresses
  13. L6:355 excites
  14. L4:325 suppresses
  15. L5:200 excites
  16. L4:129 excites
  17. L2:48 suppresses
  18. L3:16 suppresses
  19. L1:107 suppresses
  20. L8:139 excites
  21. L9:113 excites
  22. L9:889 excites
  23. L1:17 suppresses
  24. L6:133 suppresses

Feeds later layers

  1. L12:831 strongly excites
  2. L13:542 strongly excites
  3. L16:1052 strongly suppresses
  4. L12:171 strongly excites
  5. L16:483 excites
  6. L18:258 suppresses
  7. L13:627 suppresses
  8. L18:608 excites
  9. L16:693 excites
  10. L15:1058 suppresses
  11. L19:565 excites
  12. L16:612 suppresses
  13. L16:409 excites
  14. L16:443 excites
  15. L12:590 suppresses
  16. L12:293 excites
  17. L16:841 suppresses
  18. L16:643 excites
  19. L18:199 excites
  20. L18:688 excites
  21. L17:331 excites
  22. L12:1217 excites
  23. L15:140 suppresses
  24. L19:469 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 -2.78 and -2.641 values between -1.83 and -1.702 values between -1.70 and -1.563 values between -1.56 and -1.427 values between -1.42 and -1.298 values between -1.29 and -1.158 values between -1.15 and -1.0219 values between -1.02 and -0.8845 values between -0.88 and -0.7579 values between -0.75 and -0.61151 values between -0.61 and -0.48248 values between -0.48 and -0.34440 values between -0.34 and -0.21866 values between -0.21 and -0.071417 values between -0.07 and 0.06336 values between 0.06 and 0.20121 values between 0.20 and 0.3448 values between 0.34 and 0.4732 values between 0.47 and 0.6119 values between 0.61 and 0.7414 values between 0.74 and 0.8812 values between 0.88 and 1.014 values between 1.01 and 1.154 values between 1.15 and 1.284 values between 1.28 and 1.426 values between 1.42 and 1.555 values between 1.55 and 1.693 values between 1.69 and 1.821 values between 1.82 and 1.962 values between 2.09 and 2.231 values between 2.23 and 2.361 values between 2.50 and 2.64offnegativepositiveoften

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. 호
  2. ổi
  3. ạng
  4. intosh
  5. semb
  6. SCN
  7. Donetsk
  8. ло
  9. -supported
  10. izzle

Suppresses

  1. ?’
  2. ”—
  3. .—
  4. )”
  5. !’
  6. 叼
  7. ’es
  8. )—
  9. ?”
  10. ожно