Neuron Explorer

Qwen3-0.6B

"of" after a noun phrase

A neuron in layer 6 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 responds to the word "of" that follows a noun, such as "the source region of" or "the starting point of." It does not clearly separate the strong and weak cases, and turning it up did not give a clear theme.

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

    is in the forefront of the subject matter taught in schools. Whether or not these institutions have made a beneficial impact on education remains to be seen.⏎Gone are the days when mom and dad sent their children to school with the expectation for them to

  2. #2 peak

    the Aleutian low is associated with smaller eastward-moving low- and high-pressure centres, the region’s average pressure is low. It is the source region of the maritime polar air masses that influence the climate of the western U.S.

  3. #3 peak

    /wiki/Cookie<|endoftext|>Search through our library of courses, select a category you might be interested in, use the search bar, or filter options on the left side of the page.⏎Are you looking for a unique way to teach storytelling elements to your

  4. #4 peak

    blending of utility and⏎iconicity typical of the artifacts at this site.⏎As far as this author is aware, this is the first discovery in the western hemisphere of an ancient artifact in the classic Upper Palaeolithic Venus form.⏎Of course it

  5. #5 peak

    translation). The fundamental question was which of the two processes, i.e. transcription or translation, plays the dominant role in regulating cellular protein levels.⏎The starting point of the MDC researchers was to measure the turnover of cellular mRNAs and proteins

  6. #6 peak

    1.3 percent in April.⏎In this post, we take a closer look at the sources of this acceleration in prices. We divide the “basket” of goods and services that make up the CPI into twenty-three items—such as medical

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: a new type of animal called a... a type of... (plant) called a... (plant) called a... (plant) called a... (plant) called a... (plant) called a... (plant) called a...

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

    than those who ate it once in a while. The study, published in the Archives of Internal Medicine, compared the diet, chocolate intake, and body mass index of over 1,000 men and women. The researchers used BMI as

  2. #2 peak

    conducted at the Oregon National Primate Research Center at Oregon Health & Science University, monitored the reproductive systems of healthy female nonhuman primates following exposure to Delta-9-tetrahydro

  3. #3 peak

    in and pect<|endoftext|>Bacterial whole-cell biosensors were used to measure the concentration of chlortetracycline (CTC) in the feces of pigs. In this study, the Escherichia coli biosensor used has

  4. #4 peak

    tough times. As a result, scientists believe that early humans who were co-ordinated and rhythmic could have had an evolutionary advantage.⏎The researchers examined the DNA of a group of dancers and non-dancers and found that the dancers all shared two

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. L4:1686 excites
  2. L4:1966 suppresses
  3. L4:619 weakly suppresses
  4. L5:363 weakly excites
  5. L4:2646 weakly excites
  6. L1:1763 weakly excites
  7. L4:1445 weakly excites
  8. L2:277 weakly excites
  9. L5:1808 weakly excites
  10. L2:273 weakly excites
  11. L0:2243 weakly excites
  12. L5:2977 weakly excites
  13. L5:96 weakly suppresses
  14. L5:75 weakly excites
  15. L0:2828 weakly excites
  16. L4:1297 weakly suppresses
  17. L1:2976 weakly excites
  18. L5:1272 weakly excites
  19. L1:1833 weakly suppresses
  20. L5:324 weakly excites
  21. L3:2887 weakly excites
  22. L5:1797 weakly excites
  23. L5:2611 weakly excites
  24. L3:2230 weakly suppresses

Feeds later layers

  1. L11:3015 suppresses
  2. L18:415 suppresses
  3. L12:1653 suppresses
  4. L12:1518 suppresses
  5. L17:865 suppresses
  6. L12:1904 suppresses
  7. L10:159 excites
  8. L7:743 suppresses
  9. L12:2849 suppresses '-based' after a place name
  10. L13:1359 suppresses
  11. L16:1934 suppresses
  12. L7:304 suppresses
  13. L15:2633 suppresses
  14. L9:2587 suppresses
  15. L15:2700 suppresses
  16. L11:361 suppresses
  17. L17:2749 suppresses
  18. L7:194 suppresses
  19. L12:2535 suppresses
  20. L13:554 suppresses 'of' linking a noun to a larger whole
  21. L12:1286 suppresses
  22. L17:1888 weakly suppresses
  23. L16:2852 weakly suppresses 'in' after a noun or participle
  24. L10:362 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 -6.43 and -6.231 values between -5.20 and -4.991 values between -4.99 and -4.791 values between -3.76 and -3.562 values between -2.74 and -2.531 values between -2.33 and -2.121 values between -2.12 and -1.922 values between -1.92 and -1.711 values between -1.71 and -1.515 values between -1.51 and -1.306 values between -1.30 and -1.108 values between -1.10 and -0.8922 values between -0.89 and -0.6922 values between -0.69 and -0.48160 values between -0.48 and -0.281081 values between -0.28 and -0.072236 values between -0.07 and 0.14288 values between 0.14 and 0.3443 values between 0.34 and 0.5514 values between 0.55 and 0.757 values between 0.75 and 0.964 values between 1.16 and 1.371 values between 1.57 and 1.78offnegativepositiveoften

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. 诛
  2. ((-
  3. 知識
  4. 蒌
  5. .sy
  6. ivel
  7. 墅
  8. nuit
  9. 纪
  10. 把这些

Suppresses

  1. ipeg
  2. itioner
  3. emma
  4. JSGlobal
  5. anon
  6. セル
  7. aclass
  8. lod
  9. เย
  10. oplast