Neuron Explorer

Qwen3-0.6B

'Sub' starting a word or name

A neuron in layer 25 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 syllable 'Sub' at the start of a word or name, such as Subrahmanya, Subscribing, Subhash or SUBTRACTION. Turning it down made the model write words like 'subpar' and 'sub sub', so it seems tied to that spelling.

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

    to his native place Maṇali. Once he went to Tiruttaṇi, a nearby sacred place and practiced special spiritual disciplines, meditating on the deity Subrahmaṇya. A strange mystical experience here induced in him a remarkable power

  2. #2 peak

    . As the resulting jets of gas shoot backward, the engine and its craft are thrust forward. However, showing you is much cooler than explaining it.⏎Thanks for Subscribing!⏎You'll need to confirm your email, so please check your inbox

  3. #3 peak

    as Muttusvāmi Diksitar or just ‘Diksitar’. He was the son of Rāmasvāmi Diksitar and Subbalakṣmi Ammāl, born as a result of the grace

  4. #4 peak

    |⏎Koothus in Silapathikaram⏎by Subashini Pathmanathan⏎Ancient Tamils called their dances by the generic

  5. #5 peak

    aspirin-related products or acetaminophen which can react dangerously with prescription medications.⏎Treating Allergies When You Have Fibromyalgia⏎Dr. Sublett

  6. #6 peak

    , unfortunately none of which were investigated thoroughly by successive governments; leaving people in the dark about one of the most beloved leaders India has ever produced.⏎Life history of Subhash Chandra Bose⏎Subhash Chandra Bose was born on 23

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: let's take a moment to let go of the moment of momentale (momentale is the moment of momentale, momentale, momentale, momentale, momentale...). Let's take a moment to let go of the moment

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

    Product designing plays an important role in product selling. The company whose product is better and attractively designed sells more than the product of a company whose design happens to be weak and unattractive.⏎In this way, it can be said that the

  2. #2 peak

    performance of an app's launch time of less than one second to be excellent, less than 5 seconds to be good, and greater than 5 seconds to be poor.⏎Sometimes you also have to consider other metrics that don't have a direct

  3. #3 peak

    the tumour during surgery. If cancer cells are found at the edges (outer margins) of this tissue, it is called a positive surgical margin and the prognosis is less favourable. If cancer cells are not found there

  4. #4 peak

    have the most favourable outcome. Tumours that are in later stages (they have spread into nearby lymph nodes or organs far from the bile ducts) have a less favourable prognosis.⏎Location of the tumour @(Model.HeadingTag)>⏎T

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. L21:1246 excites
  2. L22:151 suppresses
  3. L19:2331 weakly suppresses
  4. L20:796 weakly excites
  5. L8:12 weakly excites
  6. L24:105 weakly suppresses
  7. L1:1528 weakly suppresses
  8. L18:81 weakly suppresses
  9. L0:2877 weakly suppresses
  10. L9:464 weakly suppresses
  11. L18:1805 weakly suppresses
  12. L14:1129 weakly suppresses
  13. L21:3033 weakly suppresses
  14. L22:400 weakly excites
  15. L17:515 weakly suppresses
  16. L21:2182 weakly excites
  17. L15:1361 weakly excites
  18. L23:143 weakly suppresses
  19. L3:2297 weakly suppresses
  20. L21:1214 weakly excites
  21. L17:3063 weakly suppresses
  22. L21:271 weakly suppresses
  23. L0:2546 weakly suppresses
  24. L20:287 weakly suppresses

Feeds later layers

  1. L27:1001 strongly suppresses
  2. L26:2794 weakly suppresses
  3. L27:1156 weakly suppresses
  4. L27:908 weakly suppresses
  5. L27:1128 weakly suppresses
  6. L27:1107 weakly suppresses
  7. L27:903 weakly suppresses
  8. L26:3069 weakly suppresses
  9. L27:364 weakly suppresses
  10. L26:2062 weakly suppresses
  11. L26:1462 weakly excites
  12. L26:1436 weakly suppresses
  13. L27:1889 weakly suppresses
  14. L26:1800 weakly suppresses
  15. L26:1962 weakly suppresses
  16. L26:1746 weakly excites
  17. L26:2906 weakly excites
  18. L27:1944 weakly suppresses
  19. L27:2818 weakly excites
  20. L26:2228 weakly excites
  21. L27:945 weakly suppresses
  22. L26:1845 weakly suppresses
  23. L26:1833 weakly excites
  24. L26:1000 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.

3 values between -17.88 and -16.371 values between -13.37 and -11.862 values between -11.86 and -10.362 values between -10.36 and -8.863 values between -8.86 and -7.359 values between -7.35 and -5.854 values between -5.85 and -4.3515 values between -4.35 and -2.8435 values between -2.84 and -1.342853 values between -1.34 and 0.16926 values between 0.16 and 1.6718 values between 1.67 and 3.1710 values between 3.17 and 4.677 values between 4.67 and 6.181 values between 6.18 and 7.683 values between 7.68 and 9.183 values between 9.18 and 10.683 values between 10.68 and 12.191 values between 12.19 and 13.692 values between 13.69 and 15.191 values between 15.19 and 16.702 values between 16.70 and 18.201 values between 21.21 and 22.711 values between 22.71 and 24.211 values between 37.74 and 39.241 values between 40.75 and 42.25offnegativepositiveoften

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. Webpack
  2. cro
  3. ve
  4. ynet
  5. amientos
  6. vier
  7. Stre
  8. ieres
  9. éta
  10. igr

Suppresses

  1. sub
  2. sub
  3. Sub
  4. Sub
  5. _sub
  6. .sub
  7. 子
  8. =sub
  9. (sub
  10. →sub