Neuron Explorer

Qwen3-0.6B

preposition after a state word

A neuron in layer 24 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 a small preposition such as 'in', 'at' or 'of' when it follows a word describing a state, as in 'stay in darkness', 'are at rest' or 'in good health'. Turning it up made the model repeat odd words like 'hushed' and 'hibachi', so the steering result is not clearly related.

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

    affected is in our brain - our circadian rhythms.⏎Although the urge to hibernate during the winter months may be strong, humans aren't really meant to stay in darkness for months at a time. We need sunlight - if we get too little

  2. #2 peak

    . Currently, it has been checked up to the power 4.6 ⋅<|endoftext|>Picture Yourself in Orbit: Are You Really Weightless?⏎An object in orbit is constantly falling, and falling is what causes “weightlessness.” Gravity acts

  3. #3 peak

    to the imbalances in the mind, spirit, and the body. If these elements are in harmony, you will remain in good health. However, if there is something that

  4. #4 peak

    nears the end of one of ...Read More⏎Uneven rainfall across the state has helped replenish Northern California reservoirs, while those to the south remain in mostly abysmal shape. Meantime, the latest snow report is in

  5. #5 peak

    was twice as frequent in those who had intermediate cardiovascular health and two and a half times more frequent in those who had poor cardiovascular health when compared to those who were in good health. With anxiety disorder, the frequency of depression was 20%

  6. #6 peak

    destroy tissue and release energy, thereby generating heat. Our metabolism speeds up when we exert ourselves, and we burn more calories. Our metabolism slows down when we are at rest, and we burn fewer calories. But even at rest we continue to expend

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: [t]d endt endt endt endt endt endt endt endt endt endt endt endt endt endt endt endt endt endt endt endt endt endt

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

    It may cause you to feel sleepy, especially when you first start taking it.⏎|Type of medicine||An antiepileptic medicine|⏎|Available as||Tablets and oral liquid medicine (suspension)|⏎Having epilepsy means that you

  2. #2 peak

    ) is a proto-oncogene that encodes a protein known as hepatocyte growth factor receptor (HG

  3. #3 peak

    ators. The less time patients spend in there the better off they are.”⏎For patients who spend several weeks in the ICU, that race is sometimes lost to late-onset sepsis, a condition where patients succumb to a systemic infection that

  4. #4 peak

    not.⏎Taken from…. The Future of Education: Tablets vs. Textbooks http://mashable.com/2012/10/05/tablets-vs-textbooks/⏎They also say…. ‘ Students with special learning needs may

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. L21:1577 excites
  2. L7:2333 weakly excites
  3. L14:2929 weakly excites
  4. L15:686 weakly suppresses
  5. L1:44 weakly excites
  6. L9:2606 weakly suppresses
  7. L1:624 weakly suppresses
  8. L15:1842 weakly excites
  9. L23:1294 weakly excites
  10. L21:1511 weakly suppresses
  11. L22:2908 weakly suppresses
  12. L2:1109 weakly suppresses
  13. L7:1288 weakly excites
  14. L23:1343 weakly suppresses
  15. L22:2377 weakly excites
  16. L8:2095 weakly suppresses
  17. L17:278 weakly excites 'as' in 'as well as' and 'as much as'
  18. L8:1846 weakly suppresses
  19. L2:570 weakly suppresses
  20. L18:2425 weakly excites
  21. L11:1906 weakly excites
  22. L0:1217 weakly excites
  23. L15:843 weakly excites
  24. L15:1951 weakly excites

Feeds later layers

  1. L27:2112 weakly excites
  2. L26:1833 weakly excites
  3. L27:2899 weakly excites
  4. L25:1591 weakly excites
  5. L27:654 weakly suppresses
  6. L25:2473 weakly excites
  7. L25:976 weakly excites
  8. L26:977 weakly suppresses
  9. L26:1881 weakly excites
  10. L25:2654 weakly excites
  11. L27:2441 weakly excites
  12. L25:1262 weakly excites
  13. L27:2098 weakly excites
  14. L25:1924 weakly excites
  15. L25:1940 weakly excites
  16. L27:2614 weakly excites
  17. L25:1752 weakly excites
  18. L27:1164 weakly excites
  19. L26:269 weakly excites
  20. L25:1881 weakly excites
  21. L25:1005 weakly excites
  22. L27:2488 weakly excites
  23. L27:2916 weakly excites
  24. L25:1913 weakly excites 'mil' or 'bil' inside names

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 -17.63 and -16.731 values between -16.73 and -15.832 values between -13.13 and -12.232 values between -12.23 and -11.331 values between -11.33 and -10.432 values between -9.53 and -8.634 values between -8.63 and -7.736 values between -7.73 and -6.8310 values between -6.83 and -5.9310 values between -5.93 and -5.037 values between -5.03 and -4.1315 values between -4.13 and -3.2328 values between -3.23 and -2.3352 values between -2.33 and -1.43210 values between -1.43 and -0.532695 values between -0.53 and 0.37612 values between 0.37 and 1.2785 values between 1.27 and 2.1755 values between 2.17 and 3.0725 values between 3.07 and 3.9724 values between 3.97 and 4.8715 values between 4.87 and 5.778 values between 5.77 and 6.6610 values between 6.66 and 7.566 values between 7.56 and 8.464 values between 8.46 and 9.364 values between 9.36 and 10.262 values between 10.26 and 11.163 values between 11.16 and 12.062 values between 12.06 and 12.962 values between 13.86 and 14.763 values between 14.76 and 15.661 values between 15.66 and 16.561 values between 17.46 and 18.36offnegativepositiveoften

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. 失
  2. rem
  3. balance
  4. BAL
  5. bal
  6. Balancer
  7. 平衡
  8. memo
  9. balance
  10. 開放

Suppresses

  1. tin
  2. t
  3. ニング
  4. olist
  5. etting
  6. panion
  7. inated
  8. /xhtml
  9. àn
  10. atics