Neuron Explorer

Qwen3-0.6B

opening parenthesis after a term

A neuron in layer 15 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 wakes up on an opening bracket that comes right after a name or term, where the text is about to give a short form or another name for it, such as "International Monetary Fund (IMF)". Turning it up did not produce a clear change in what the model wrote.

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

    Financial Protection Bureau (CFPB) released a study reviewing financial education research priorities for America’s youth.⏎Congressmen Bill Foster (D-IL) and French Hill (R-AR) both introduced H. Res. 327, which

  2. #2 peak

    variety of cultivated vegetables and wild greens are eaten as well as bamboo shoots. Meals often are accompanied by lentils, pickled relishes, and balachaung (made from fried dry prawns). There are a variety of

  3. #3 peak

    the sergeants and corporals were to be named by the captains. Those who participated in these organizations would be exempt from service as couriers and cordilleras (rangers), and their horses would be exempt from seizure for auxiliaries

  4. #4 peak

    percentile value) were associated with an increased risk of diarrheal disease (incidence rate ratio = 1.39, 95% confidence interval: 1.03, 1.87), but only after a

  5. #5 peak

    over a period of 15 years (1996 to 2011) with different levels of harvesting intensity, including selective harvesting at low (13.0% removal of growing stock volume), medium (29.

  6. #6 peak

    a funding source for Greece.⏎On 2 May 2010, the Eurozone countries, European Central Bank (ECB) and International Monetary Fund (IMF), later nicknamed the Tro

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 other side, the other side of the other side, the other side of the other side of the other... and of the of of of of... the of the of of of of of... of of of... of of... 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

    demonstration. It took about 10 minutes to fire up the steam pump to 140 pounds pressure. A cistern buried in the intersection (placed there for fire protection) provided the source of water. Using three hoses at once

  2. #2 peak

    resemble the wood of Olive trees in the Olea genus. And it should come as little surprise that Olive Ash can be a dead ringer for actual Olive (with the exception of the porous grain structure, which gives its true identity away easily),

  3. #3 peak

    Centre of Excellence in Quantum Computing.<|endoftext|>In ancient China one’s Zodiac sign was fixed by the position of the planet Jupiter during the year of one’s birth (it takes twelve years for Jupiter to orbit the sun). On this level, Ai Wei

  4. #4 peak

    cause of equine ulcers. The lifestyle of many horses, which includes confinement, high level competition, high energy diets rather than pasture, contributes to stress (and ulcers). So does any significant change in the horse’s routine or life style

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. L14:2872 strongly suppresses
  2. L12:2745 excites
  3. L6:2439 excites
  4. L14:880 suppresses
  5. L14:166 excites
  6. L14:2752 excites
  7. L11:50 excites
  8. L10:66 suppresses opening parenthesis after a citation or term
  9. L12:1807 suppresses
  10. L0:1978 excites
  11. L14:1548 suppresses
  12. L10:1580 excites
  13. L11:1579 suppresses
  14. L13:1078 excites
  15. L14:2066 excites
  16. L7:419 excites
  17. L14:43 weakly suppresses
  18. L11:2689 weakly excites
  19. L10:242 weakly excites
  20. L14:1209 weakly excites
  21. L13:840 weakly excites
  22. L14:1526 weakly excites
  23. L9:2259 weakly suppresses
  24. L13:2609 weakly excites

Feeds later layers

  1. L17:933 suppresses
  2. L16:634 weakly excites
  3. L18:299 weakly suppresses
  4. L16:882 weakly suppresses
  5. L17:617 weakly excites
  6. L19:911 weakly excites
  7. L16:2250 weakly suppresses
  8. L18:1403 weakly suppresses
  9. L18:83 weakly excites
  10. L16:2251 weakly suppresses
  11. L16:1977 weakly suppresses
  12. L21:2393 weakly suppresses
  13. L22:592 weakly excites
  14. L18:1174 weakly suppresses
  15. L19:871 weakly suppresses
  16. L16:2018 weakly suppresses
  17. L21:1341 weakly suppresses
  18. L16:565 weakly excites
  19. L18:3022 weakly suppresses
  20. L18:2651 weakly suppresses
  21. L19:1982 weakly excites
  22. L16:2048 weakly suppresses
  23. L19:376 weakly suppresses
  24. L19:2060 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 -2.81 and -2.671 values between -2.53 and -2.392 values between -1.69 and -1.553 values between -1.55 and -1.414 values between -1.41 and -1.273 values between -1.27 and -1.131 values between -1.13 and -0.997 values between -0.99 and -0.858 values between -0.85 and -0.7220 values between -0.72 and -0.5834 values between -0.58 and -0.44105 values between -0.44 and -0.30300 values between -0.30 and -0.161199 values between -0.16 and -0.021769 values between -0.02 and 0.12277 values between 0.12 and 0.2673 values between 0.26 and 0.4045 values between 0.40 and 0.5421 values between 0.54 and 0.686 values between 0.68 and 0.828 values between 0.82 and 0.965 values between 0.96 and 1.105 values between 1.10 and 1.243 values between 1.24 and 1.383 values between 1.38 and 1.522 values between 1.66 and 1.801 values between 1.80 and 1.941 values between 1.94 and 2.071 values between 2.63 and 2.77offnegativepositiveoften

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. -Mart
  3. -Se
  4. --*/⏎
  5. 认识到
  6. -drive
  7. dı
  8. ----------- ⏎
  9. ivery
  10. -labelledby

Suppresses

  1. bish
  2. lore
  3. yes
  4. AGIC
  5. yeah
  6. lotte
  7. .ac
  8. cce
  9. yog
  10. YES