Neuron Explorer

Qwen3-0.6B

'in' after a noun

A neuron in layer 17 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

The neuron lights up on the word 'in' when it follows a noun and adds a location or context to it, like 'the energy released in these collisions' or 'the options in the drop-down menu'. Turning it up made the writing repetitive and off-track, so the steering does not show the pattern clearly.

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

    way down these particles are slowed down by Earth’s atmosphere, which acts as a shield. These charged particles collide with gas molecules in the atmosphere. The energy released in these collisions is given off as light.⏎Geomagnetic storms follow the 11

  2. #2 peak

    to use technical jargon as a means to succintly describe the work done. However, if the target audience does not have sufficient understanding of the subject included in the report, care must be taken in order to avoid or prevent the use of

  3. #3 peak

    you’re seeing in the clip above:⏎Some of the stellar debris falls into the black hole and some of it is ejected into space at high speeds. The areas in white are regions of highest density, with progressively redder colors corresponding to lower-density

  4. #4 peak

    earnings is substantial. If a low value-added teacher in the bottom 5% is replaced by an average teacher, not even an excellent one, the expected increase in the present value of students lifetime income is more than $250,0

  5. #5 peak

    to the Zipped file from the File Explorer and right-click on it. From the options in the drop-down menu, choose Extract All.⏎Step 2: A folder with

  6. #6 peak

    the country by the beginning of the 20th century, it was New Orleans' version that had the most influence on Chicago jazz and early swing that followed in the 1920s. Drawing on a thorough and imaginative command of

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 very big a large set of a very large set of a large set of a large set of a large set of a large set of a large set of a large set of a large set of a large set of a large set of

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

    story plausible, we can argue that, among other things, Western countries uphold conceptions of nationalism that center around individual rights and

  2. #2 peak

    least acidic among citrus fruits, but it has many seeds. The rind is thin and mand

  3. #3 peak

    in finding the best solution from a large number of options.⏎They found that classical methods still performed best in these

  4. #4 peak

    humanity and spark global change.⏎The Tech Awards, a signature program of The Tech Museum of Innovation in San Jose, California, selected the Drift Catcher from among hundreds of nominations representing 68 countries, for "empowering communities for health

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. L13:132 excites "of" after a noun
  2. L16:1807 suppresses 'in' after a motion verb
  3. L10:132 excites
  4. L5:763 suppresses
  5. L8:45 suppresses
  6. L3:232 excites 'on' after a noun like 'impact on'
  7. L4:4 suppresses
  8. L8:706 suppresses
  9. L15:906 suppresses
  10. L12:1467 weakly excites numbers after dates and figures
  11. L5:729 weakly suppresses
  12. L14:539 weakly excites
  13. L4:1035 weakly suppresses 'for' after verbs like 'fall' or 'click'
  14. L10:469 weakly suppresses 'of' after a noun phrase
  15. L10:272 weakly suppresses
  16. L4:1500 weakly suppresses
  17. L8:3016 weakly excites
  18. L14:917 weakly excites
  19. L14:1519 weakly excites
  20. L4:192 weakly excites
  21. L15:894 weakly excites
  22. L14:1326 weakly suppresses
  23. L4:1966 weakly suppresses
  24. L3:311 weakly suppresses

Feeds later layers

  1. L18:1173 suppresses
  2. L20:880 suppresses
  3. L20:1494 weakly suppresses
  4. L21:1061 weakly suppresses
  5. L19:1724 weakly suppresses
  6. L22:1524 weakly suppresses
  7. L21:1030 weakly excites
  8. L19:7 weakly suppresses
  9. L22:194 weakly excites
  10. L22:103 weakly excites
  11. L18:2384 weakly excites
  12. L19:2470 weakly suppresses
  13. L20:1423 weakly excites
  14. L20:47 weakly suppresses
  15. L19:914 weakly excites
  16. L26:284 weakly excites
  17. L27:797 weakly suppresses
  18. L19:532 weakly excites
  19. L19:769 weakly suppresses
  20. L27:2106 weakly suppresses
  21. L19:2425 weakly suppresses
  22. L18:604 weakly suppresses
  23. L26:1959 weakly suppresses
  24. L19:219 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 -11.13 and -10.701 values between -8.98 and -8.551 values between -7.26 and -6.831 values between -6.83 and -6.401 values between -6.40 and -5.971 values between -5.54 and -5.112 values between -5.11 and -4.687 values between -4.25 and -3.826 values between -3.82 and -3.398 values between -3.39 and -2.965 values between -2.96 and -2.5310 values between -2.53 and -2.1017 values between -2.10 and -1.6731 values between -1.67 and -1.2448 values between -1.24 and -0.81169 values between -0.81 and -0.382030 values between -0.38 and 0.051294 values between 0.05 and 0.48144 values between 0.48 and 0.9142 values between 0.91 and 1.3427 values between 1.34 and 1.7724 values between 1.77 and 2.2013 values between 2.20 and 2.636 values between 2.63 and 3.063 values between 3.06 and 3.485 values between 3.48 and 3.915 values between 3.91 and 4.342 values between 4.77 and 5.201 values between 5.20 and 5.633 values between 5.63 and 6.06offnegativepositiveoften

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. licity
  2. venth
  3. чат
  4. ധ
  5. usch
  6. harga
  7. since
  8. ícia
  9. izzazione
  10. irth

Suppresses

  1. different
  2. 不同的
  3. different
  4. Gos
  5. ographs
  6. .equalTo
  7. 此基础上
  8. 多样化
  9. 完全不同
  10. sel