Neuron Explorer

Qwen3-0.6B

gap right before a number

A neuron in layer 25 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 spot just before a number is expected, such as after 'over the age of' or 'kids under', where the next word will be a figure. Turning it up made the text repetitive and unfocused. It is weaker and less clear on the opposite side.

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

    in the two collaborating project laboratories. This work is<|endoftext|>“Holy Week: In Like a Lion, Out Like a Lamb” (Matthew 21:1-11; 27:11-66)⏎“In like a

  2. #2 peak

    23 types of pneumococcal bacteria that can lead to pneumonia, blood-stream infections and meningitis. A shot is recommended for everyone over the age of 65 and for those under 65 who have the same risk

  3. #3 peak

    bacteria. The disease can⏎affect the heart, joints, skin and brain.⏎Children in the age⏎group of 5 to 15 years are easy prey to this disease. The disease develops

  4. #4 peak

    project is eligible for School of Mathematics Doctoral Training Grant funding - please contact us for more information.<|endoftext|>História (São Paulo)⏎On-line version ISSN 1980-4369⏎DOMINGUES, Beatriz

  5. #5 peak

    Richeson, and such an appliance is utilized to expand that arch. The palatal expander is best used while a child is still growing, usually between ages 8 and 10.⏎"By age 7, a child will have

  6. #6 peak

    .⏎Yüksek Lisans Öğrenimine MSÜ Barbaros DEBİM Enstitüsünde devam etmekte aynı zamanda enerji sistem ve modelleme üzerine araştırmalar yapmaktadır.<|endoftext|>-

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 key to a room even roomy even roomy even roomy even roomy even roomy even roomy even roomy even roomy even roomy even roomy even roomy even roomy even roomy even roomy even

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

    ched, soft rays. The anal fin has one spiny ray and 24-26 soft rays. The caudal fin is truncated.⏎The upper part of the body is yellowish or brownish, the lower part is white

  2. #2 peak

    zoom in or out). Your responses should demonstrate critical thinking. Provide justification for your analysis:⏎How many earthquakes are listed on this map?⏎How many earthquakes are over 6.0 over the week? How many between 4.5 and

  3. #3 peak

    (Hindi:कामायनी) (1936) is a Hindi epic poem (Mahakavya) by Jaishankar Prasad (1889–1937). It is

  4. #4 peak

    question for which a “yes/no” is perfectly adequate. This is not the case with a feature that has a specific tolerance i.e.: a nominal and an upper and lower spec limit (1.0 +/- .005). It might

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. L20:54 excites
  2. L20:3033 suppresses
  3. L19:305 weakly excites
  4. L22:349 weakly excites
  5. L3:1179 weakly excites
  6. L0:178 weakly suppresses
  7. L18:251 weakly excites
  8. L18:125 weakly suppresses
  9. L20:541 weakly excites
  10. L19:859 weakly excites
  11. L0:1301 weakly suppresses
  12. L0:1224 weakly suppresses
  13. L15:184 weakly excites
  14. L17:292 weakly excites
  15. L3:283 weakly excites
  16. L18:178 weakly excites
  17. L0:2792 weakly suppresses middle pieces of split words
  18. L20:270 weakly excites
  19. L17:17 weakly suppresses
  20. L14:1289 weakly excites
  21. L24:685 weakly suppresses
  22. L7:92 weakly excites
  23. L14:20 weakly excites
  24. L11:844 weakly excites

Feeds later layers

  1. L27:1164 weakly excites
  2. L26:1482 weakly excites
  3. L27:1985 weakly excites
  4. L26:1952 weakly excites
  5. L26:1029 weakly excites
  6. L26:637 weakly excites
  7. L27:2381 weakly excites
  8. L27:1846 weakly excites
  9. L27:848 weakly excites
  10. L26:1213 weakly excites
  11. L27:2131 weakly suppresses
  12. L27:1690 weakly excites
  13. L26:353 weakly suppresses
  14. L26:2969 weakly excites
  15. L27:285 weakly suppresses
  16. L27:837 weakly excites
  17. L27:1410 weakly excites "as well" and "as much as"
  18. L27:14 weakly excites
  19. L26:2865 weakly excites
  20. L27:2218 weakly excites
  21. L27:893 weakly suppresses
  22. L26:1509 weakly suppresses
  23. L27:2234 weakly suppresses
  24. L26:1356 weakly 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 -9.74 and -9.312 values between -9.31 and -8.891 values between -7.17 and -6.741 values between -6.32 and -5.893 values between -5.89 and -5.461 values between -5.46 and -5.031 values between -5.03 and -4.603 values between -4.60 and -4.173 values between -4.17 and -3.753 values between -3.75 and -3.322 values between -3.32 and -2.898 values between -2.89 and -2.467 values between -2.46 and -2.0314 values between -2.03 and -1.6016 values between -1.60 and -1.1865 values between -1.18 and -0.75467 values between -0.75 and -0.322063 values between -0.32 and 0.11995 values between 0.11 and 0.54187 values between 0.54 and 0.9732 values between 0.97 and 1.3914 values between 1.39 and 1.827 values between 1.82 and 2.252 values between 2.25 and 2.683 values between 2.68 and 3.111 values between 3.11 and 3.543 values between 4.39 and 4.821 values between 4.82 and 5.251 values between 5.68 and 6.111 values between 6.96 and 7.39offnegativepositiveoften

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. ise
  2. 娩
  3. wort
  4. unta
  5. gne
  6. êt
  7. azine
  8. ahoo
  9. ServerError
  10. wj

Suppresses

  1. setDefault
  2. cí
  3. pery
  4. ума
  5. os
  6. even
  7. =os
  8. chimp
  9. -DD
  10. hers