Neuron Explorer

Qwen3-0.6B

unclear: syllables inside Sanskrit religious words

A neuron in layer 22 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron lights up on a fragment in the middle of a Hindi or Sanskrit-derived word, such as the 'ak' in Mahakavya or the 'ad' in Ekadasi. It is not clear what ties these together, and turning it up made the model's text wander and repeat.

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

    cooperation.<|endoftext|>Kamayani (Hindi:कामायनी) (1936) is a Hindi epic poem (Mahakavya) by Jaishankar Prasad (1889–

  2. #2 peak

    and form a halo around her. In front of the goddess are three Mahakalas (enlightened beings who protect those on the path to Buddhah

  3. #3 peak

    of the Buddhist deity Chakrasamvara⏎- The Buddhist deity Mahakala as a Brahman⏎- The Buddhist deity Simhavaktra Dak

  4. #4 peak

    n “Souramana” during the auspicious month of Dhanurmasa (Margasira Poushya), out of 24 Ekadasi Tidhis in a year Suklapaksha Ekadasi is termed

  5. #5 peak

    .⏎Metsäpohjainen piha paransi päiväkotilasten immuunijärjestelmää jo kuukaudessa<|endoftext|>Research Looks at Alternative Power for Military⏎Source Newsroom: South

  6. #6 peak

    on this day it is further called “Mukkoti Ekadasi”.⏎It is even named as Vaikuntha Ekadasi as

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: 50000000000000000000000000000000000000000000000

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

    , which means to combine. Our ultimate goal is union with our eternal, transcendent self.⏎There are many schools of thought on yoga. Some of them contradict one another. There are several schools of thought about yoga. please click the up coming

  2. #2 peak

    the trees and sang so sweetly that the children used to stop their games in order to listen to them. ‘How happy we are here!’ they cried to each other.⏎One day the Giant came back. He had been to visit his friend

  3. #3 peak

    objective nature of reality, but primarily predicated on the the fact that humans subjectively interpret the world in terms of objects (vertices) and their respective relationships to one another (edges) (an argument against this idea). The popular data model used

  4. #4 peak

    studied classical Indian texts such as the Dharmaśāstra and inscriptions make reference to Sanskrit literature. Eventually, while the Cham and Sanskrit languages influenced one another, Cham culture assimilated Hinduism, and Chams were eventually able adequately

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. L20:2919 suppresses
  2. L20:952 suppresses
  3. L19:2162 excites
  4. L21:1816 excites
  5. L21:1049 suppresses
  6. L21:2134 excites
  7. L21:2296 excites
  8. L21:1226 suppresses
  9. L19:1754 excites
  10. L21:784 suppresses
  11. L20:2070 excites
  12. L19:2606 excites
  13. L19:2104 excites
  14. L20:1798 suppresses
  15. L21:2565 excites
  16. L18:2007 excites
  17. L21:1903 suppresses
  18. L21:2419 suppresses
  19. L20:1160 excites
  20. L18:2772 excites
  21. L21:1790 excites
  22. L20:1794 suppresses
  23. L19:1933 excites
  24. L21:1925 suppresses

Feeds later layers

  1. L27:425 suppresses
  2. L25:210 suppresses
  3. L25:120 suppresses
  4. L25:420 suppresses
  5. L25:477 excites
  6. L24:48 excites
  7. L25:531 excites
  8. L26:511 suppresses
  9. L27:249 suppresses
  10. L26:265 suppresses
  11. L23:197 suppresses
  12. L25:334 excites
  13. L24:221 excites
  14. L25:298 suppresses
  15. L25:98 suppresses
  16. L26:248 suppresses
  17. L23:390 suppresses
  18. L25:1114 suppresses
  19. L23:12 suppresses
  20. L25:176 suppresses
  21. L23:122 weakly excites
  22. L27:348 weakly suppresses
  23. L27:702 weakly suppresses
  24. L25:1135 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 -40.66 and -39.501 values between -31.41 and -30.251 values between -27.94 and -26.791 values between -19.85 and -18.701 values between -12.92 and -11.761 values between -8.29 and -7.142 values between -5.98 and -4.835 values between -4.83 and -3.673 values between -3.67 and -2.5132 values between -2.51 and -1.36376 values between -1.36 and -0.203394 values between -0.20 and 0.9679 values between 0.96 and 2.115 values between 2.11 and 3.273 values between 3.27 and 4.423 values between 4.42 and 5.58offnegativepositiveoften

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. arin
  2. わけではない
  3. <<⏎
  4. 就够了
  5. 以上的
  6. >" ⏎
  7. 的不同
  8. 的基础上
  9. 的有效
  10. Willis

Suppresses

  1. num
  2. ou
  3. tered
  4. num
  5. -choice
  6. OLER
  7. ECB
  8. ersonic
  9. ič
  10. ournament