Neuron Explorer

Qwen3-0.6B

religious words like faith and christ

A neuron in layer 1 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on religious words, especially 'Christ' and 'faith', and on the 'of' in phrases like 'Advent of'. It switches off, or turns negative, on the book name 'Jeremiah' just before a Bible verse number. Steering did not clearly bring out a religious theme.

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

    , freedom for the oppressed, and abundance for the poor. Therefore, the people waited for the Messiah or Christ.⏎The word Christmas, or "Christes Maesse," was first used in 1038. Until then,

  2. #2 peak

    , atheists and agnostics. Our religious diversity enriches our cultural fabric and reminds us that what binds us as one is not the tenets of our faiths, the

  3. #3 peak

    , and our right as a people to set our own course.⏎America proudly stands with people of every nation who seek to think, believe, and practice their faiths as they choose. In the years to come, my Administration will remain committed to

  4. #4 peak

    He worked a miracle. God worked a miracle so that His people could be near Him in worship.⏎Let us not miss that meaning, as we celebrate the Advent of Jesus Christ, the Light of the world (John 8:12).

  5. #5 peak

    in northeast India.⏎"It is not only for the Buddhists but for all those who believe in Buddhism. People belonging to Hindu, Buddhist, Muslim and Christian faiths participate here," said Khisa.⏎Tripura Governor Devanand Konwar said

  6. #6 peak

    the research is completed in such a context.⏎The problem is that students in the Elders Quorum for the Laveen Ward of the Church of Jesus Christ of Latter-day Saints do not participate enough in discussion or preparation for the lesson portion 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: the fact that, as a result, the result is the same as the initial condition. So, the result is the same as the initial condition. So, the result is the same as the initial condition. So, the result is the

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

    ,7⏎- Noah – Genesis 6:13⏎III. All the promises of God are for those who hear and obey His voice⏎Jeremiah 7:23 "But this is what I commanded them, saying, '

  2. #2 peak

    with him, and he with Me.⏎IV. God often rebuked the people of Israel because they did not listen to and obey His voice⏎Jeremiah 7:28 "And you shall say to them, 'This is the

  3. #3 peak

    lump of clay in forming pottery.⏎So I went down to the potter’s house, and there he was, making something on the wheel. —Jeremiah 18:3 NASB⏎For the temple, Hiram contructed

  4. #4 peak

    born? What did they come to do? (Matthew 2:2)⏎- What prophecy did they chief priests and scribes know about Jesus? (Matthew 2:3-6)⏎- How did Herod scheme against Jesus? How

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. L0:1206 strongly suppresses
  2. L0:1224 suppresses
  3. L0:1301 suppresses
  4. L0:1749 suppresses
  5. L0:178 suppresses
  6. L0:1153 excites
  7. L0:1572 weakly excites
  8. L0:1741 weakly suppresses
  9. L0:1174 weakly excites
  10. L0:598 weakly suppresses 'the' right before a year or number
  11. L0:1527 weakly suppresses
  12. L0:2521 weakly excites
  13. L0:3030 weakly excites
  14. L0:411 weakly suppresses
  15. L0:827 weakly excites
  16. L0:1525 weakly suppresses
  17. L0:3040 weakly excites
  18. L0:25 weakly excites
  19. L0:2555 weakly suppresses
  20. L0:2239 weakly excites
  21. L0:839 weakly excites
  22. L0:2642 weakly suppresses
  23. L0:2034 weakly suppresses
  24. L0:1172 weakly suppresses

Feeds later layers

  1. L2:1203 suppresses
  2. L20:2532 suppresses
  3. L22:2558 suppresses
  4. L2:290 suppresses
  5. L20:2806 suppresses
  6. L2:162 suppresses
  7. L7:371 suppresses
  8. L19:1163 suppresses
  9. L9:147 suppresses
  10. L18:3038 weakly suppresses
  11. L19:2131 weakly suppresses
  12. L2:401 weakly suppresses
  13. L23:1809 weakly suppresses
  14. L25:1599 weakly suppresses
  15. L13:2533 weakly suppresses
  16. L5:1834 weakly excites
  17. L5:1740 weakly suppresses
  18. L25:1653 weakly suppresses
  19. L6:1441 weakly suppresses
  20. L21:215 weakly excites
  21. L6:3006 weakly suppresses
  22. L3:832 weakly excites
  23. L20:979 weakly suppresses
  24. L8:2911 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.

1 values between -3.90 and -3.791 values between -3.55 and -3.441 values between -3.09 and -2.971 values between -2.97 and -2.861 values between -2.62 and -2.511 values between -2.04 and -1.921 values between -1.81 and -1.691 values between -1.58 and -1.462 values between -1.23 and -1.114 values between -0.99 and -0.881 values between -0.88 and -0.763 values between -0.76 and -0.642 values between -0.64 and -0.533 values between -0.53 and -0.417 values between -0.41 and -0.3035 values between -0.30 and -0.18341 values between -0.18 and -0.061830 values between -0.06 and 0.051175 values between 0.05 and 0.17406 values between 0.17 and 0.2979 values between 0.29 and 0.4010 values between 0.40 and 0.522 values between 0.64 and 0.75offnegativepositiveoften

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. iga
  2. inas
  3. LE
  4. isten
  5. punk
  6. �
  7. 骨干
  8. eng
  9. 另一种
  10. algebra

Suppresses

  1. Hebrew
  2. Sabbath
  3. stoi
  4. Cohen
  5. cate
  6. Heb
  7. habit
  8. Rabbi
  9. 邈
  10. Joi