Neuron Explorer

Qwen3-0.6B

'few' after 'a' in quantity phrases

A neuron in layer 2 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 word "few" when it follows "a", as in "a few hours" or "a few millimetres". When it was turned up, the model's writing did not show a clear change in 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

    body. On the other hand, the female grows to between 6 and 7cm in length and can vary in color. Mating can take from a few hours to a few days with the male sitting atop the female.<|endoftext|>Even when

  2. #2 peak

    lysozyme enzyme did not lose its activity after being released from the capsules in this way.⏎In clinical use, the laser would be able to penetrate a few millimetres of tissue. It could be shone through the skin, or

  3. #3 peak

    joy of finding what you seek doesn't last.⏎In a classic study, researchers trained monkeys to do a task and get a spinach leaf in reward. After a few days, the researchers replaced the spinach with a grape. Compared to spinach, a

  4. #4 peak

    borne illness usually can be killed when heated. Heating food to an internal temperature above 160 degrees Fahrenheit or 78 degrees Celsius, even for a few seconds, is enough to kill most bacteria, viruses, or parasites. Tox

  5. #5 peak

    foam cools and then is heated again, after which it is left out to cure for a few hours or a few days.⏎- The foam is washed, dried and inspected to

  6. #6 peak

    a collection bottle, you may cough a bit. This is because your lung re-expands to fill the space where fluid had been. This sensation lasts for a few hours after the test.⏎During the test, tell your health care provider if you

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. The first light is the 100, and the second is the 10. The system is designed such that the first is the 10. So, the system is designed such that the first is the

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

    O snap/ring gauges (for checking shafts) and plug gauges (for checking holes). They are mostly simple and easy to comprehend.⏎In the

  2. #2 peak

    given to centering the load.⏎Progressive and transfer dies produce many types of stamped parts. Consolidation of multiple parts into a single stamping and down-gauging material thickness to reduce component weight increases part complexity. This complexity often requires dies

  3. #3 peak

    fill a gap in patent scholarship, this Article offers a new framework for gauging the patentability of seemingly impossible inventions. Briefly, it contends that a

  4. #4 peak

    had gone to Arles with the intention of creating a colony of artists... but nobody came. Eventually, Paul Gauguin came to try it out.⏎Gauguin was a good friend but

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. L1:1996 weakly suppresses
  2. L0:178 weakly suppresses
  3. L0:1224 weakly suppresses
  4. L1:2474 weakly excites
  5. L1:1013 weakly suppresses
  6. L1:2848 weakly suppresses
  7. L0:1301 weakly suppresses
  8. L0:920 weakly excites digit '9' inside years and numbers
  9. L1:2801 weakly excites
  10. L1:2922 weakly excites
  11. L0:2828 weakly suppresses
  12. L1:32 weakly suppresses
  13. L0:992 weakly excites
  14. L0:414 weakly suppresses blank after a dash in number ranges
  15. L1:2599 weakly excites
  16. L0:2826 weakly excites
  17. L1:35 weakly excites
  18. L1:704 weakly suppresses
  19. L1:2093 weakly suppresses
  20. L1:24 weakly suppresses
  21. L0:690 weakly excites
  22. L1:1230 weakly excites
  23. L0:1019 weakly suppresses
  24. L1:1977 weakly suppresses

Feeds later layers

  1. L3:916 suppresses
  2. L5:2272 weakly excites
  3. L24:1620 weakly suppresses
  4. L12:1969 weakly suppresses
  5. L7:2497 weakly excites
  6. L23:1435 weakly suppresses
  7. L5:601 weakly suppresses
  8. L22:667 weakly suppresses
  9. L10:1646 weakly suppresses
  10. L3:1445 weakly suppresses
  11. L27:744 weakly suppresses
  12. L9:620 weakly excites
  13. L4:2212 weakly excites
  14. L12:2589 weakly suppresses
  15. L9:724 weakly suppresses
  16. L10:937 weakly excites
  17. L3:2371 weakly suppresses
  18. L8:1491 weakly suppresses
  19. L12:253 weakly suppresses
  20. L9:1943 weakly suppresses
  21. L8:827 weakly excites
  22. L22:496 weakly excites
  23. L23:1404 weakly excites
  24. L18:2471 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 -1.20 and -1.071 values between -1.07 and -0.942 values between -0.81 and -0.691 values between -0.69 and -0.566 values between -0.56 and -0.4314 values between -0.43 and -0.31131 values between -0.31 and -0.18856 values between -0.18 and -0.052275 values between -0.05 and 0.08543 values between 0.08 and 0.2056 values between 0.20 and 0.337 values between 0.33 and 0.461 values between 0.46 and 0.595 values between 0.84 and 0.972 values between 0.97 and 1.101 values between 1.10 and 1.221 values between 1.22 and 1.352 values between 1.61 and 1.731 values between 2.24 and 2.371 values between 3.01 and 3.131 values between 3.77 and 3.90offnegativepositiveoften

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. ертв
  2. ��
  3. 分钟后
  4. actic
  5. odied
  6. cry
  7. 医科大学
  8. elerik
  9. utorial
  10. 小时内

Suppresses

  1. 看你
  2. khổ
  3. 绞
  4. ziel
  5. ооруж
  6. 류
  7. 协
  8. устрой
  9. enkins
  10. erman