Neuron Explorer

Qwen3-0.6B

'is' after a subject noun about likelihood

A neuron in layer 21 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 'is' (or the Chinese word for rate) when a sentence says how likely, common or large something is, such as 'the chance of survival is'. Turned up, the model fell into loops of 'high' and 'low', which fits that theme of degree.

Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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

    for Thai Learners Based on the Comparative Analysis⏎- 人文学院-已发表论文 ⏎“助动词”在汉语中具有较大的作用,使用率高,但是泰国学生经常在学习和使用汉语助动词中产生

  2. #2 peak

    to the west and over 2,300 nautical miles to the south.⏎Striped marlin are found throughout the Pacific Ocean, but their abundance is much greater in the eastern half of the Pacific and appears to merge seasonally in

  3. #3 peak

    cycle. As we are heading towards the next solar maximum, due in 2013, the chance of big magnetic storms is on the increase. On average you might expect to see aurora in the far

  4. #4 peak

    also in an attempt to reduce instances of child trafficking following the tsunami, placed a temporary hold on a child adoptions.⏎In regions where the potential for child trafficking is particularly high, countries can be spurred to action following the requests of other world leaders

  5. #5 peak

    or are unable to, enter the medical system for this important screening. In Asia, the number of prospective patients that fail to enter the system for annualized screening is closer to 70%, a leading contributor for breast cancer being the number #

  6. #6 peak

    Cell Disease?⏎The symptoms of sickle cell anemia start when the babies are around 5 to 6 months old. The risk of developing new medical conditions is high when they grow older. You must consult a specialist when you notice the first

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: what are some of the useful and useful you can do in the first 5 years of your career, and what do you why you do it? Why you do it to why you do it for your future, why you do it to

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

    of individual variation, or environmental conditions.⏎You have requested a machine translation of selected content from our databases. This functionality is provided solely for your convenience and is in no way intended to replace human translation. Neither BioOne nor the owners and publishers of

  2. #2 peak

    cone,’ from the ancient Greek strobilos, meaning ‘a twisted object’, a botanical term that describes the shape of the volcano’s peak.⏎Puncak Jaya, Australia (Oce

  3. #3 peak

    Dispersal of Aedes Aegypti in an Urban Area After Blood Feeding as Demonstrated by Rubidium-Marked Eggs," American Journal of Tropical Medicine Hygiene⏎52 (2):177-17

  4. #4 peak

    and criteria as we use for PAA and CAF.⏎Through these three programs each<|endoftext|>发布时间:2017-03-27 来源:新

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. L20:1919 strongly excites
  2. L19:1454 excites
  3. L19:2023 suppresses
  4. L20:1800 suppresses
  5. L16:2072 suppresses
  6. L20:1762 suppresses
  7. L19:1165 excites
  8. L19:3017 suppresses
  9. L8:1032 suppresses
  10. L20:665 excites
  11. L19:366 suppresses
  12. L16:1721 suppresses
  13. L20:1733 weakly suppresses
  14. L18:1623 weakly excites
  15. L9:525 weakly excites
  16. L19:1915 weakly excites
  17. L17:1320 weakly excites
  18. L16:2744 weakly excites
  19. L15:2383 weakly excites
  20. L20:977 weakly suppresses
  21. L16:899 weakly excites
  22. L19:1065 weakly excites
  23. L20:1295 weakly suppresses
  24. L19:1373 weakly excites

Feeds later layers

  1. L25:934 excites
  2. L25:1993 excites
  3. L23:469 excites
  4. L22:1370 excites
  5. L23:545 excites
  6. L24:1595 excites
  7. L24:915 excites
  8. L25:461 excites
  9. L22:1407 excites
  10. L27:1258 excites
  11. L24:299 excites
  12. L23:530 excites
  13. L24:1668 excites
  14. L24:1052 excites
  15. L22:422 excites
  16. L23:1398 weakly excites
  17. L22:1046 weakly excites
  18. L25:1353 weakly excites "in" before "good" condition
  19. L25:1283 weakly excites
  20. L23:283 weakly excites
  21. L23:1266 weakly excites
  22. L22:3015 weakly excites
  23. L26:2458 weakly excites
  24. L23:514 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.

4 values between -2.90 and -2.372 values between -2.37 and -1.857 values between -1.85 and -1.32205 values between -1.32 and -0.801487 values between -0.80 and -0.271879 values between -0.27 and 0.25205 values between 0.25 and 0.7845 values between 0.78 and 1.3017 values between 1.30 and 1.8314 values between 1.83 and 2.3511 values between 2.35 and 2.887 values between 2.88 and 3.407 values between 3.40 and 3.932 values between 3.93 and 4.452 values between 4.45 and 4.983 values between 4.98 and 5.512 values between 5.51 and 6.032 values between 6.03 and 6.561 values between 7.08 and 7.611 values between 10.23 and 10.761 values between 11.28 and 11.811 values between 12.86 and 13.381 values between 14.43 and 14.962 values between 17.58 and 18.11offnegativepositiveoften

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. 很高
  2. 较高
  3. 更高
  4. 较高的
  5. 更高的
  6. 较低
  7. 很高的
  8. 高的
  9. 最高的
  10. highest

Suppresses

  1. subclass
  2. ;display
  3. diğini
  4. nc
  5. quete
  6. nek
  7. 斯塔
  8. dings
  9. iously
  10. reen