Neuron Explorer

Qwen3-0.6B

the 'per' piece in perished and per se

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 lights up on the small word piece 'per' when it starts words like 'perished', 'per se' and 'permissive'. Turning it up made the model write confused, repetitive text, so the steering results don't reveal a clear 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

    a time reinforced by soldiers, were kept at bay for many hours by two foreign burglars who defended themselves in a house with Mauser pistols, and who ultimately perished when the building caught fire and was burnt.⏎The following table shows the war

  2. #2 peak

    cohesiveness while living in a Christian world. It will be suggested that Mudejar self-perception and group-perception were founded not only on religious belief per se, but also on⏎the perpetuation of a social world the organization and

  3. #3 peak

    to the elimination of the numbered posts and the majority win feature, because as a practical matter women were no longer seen as competing against men or incumbents per se. Moreover, I see the system as giving women an excellent opportunity to organize

  4. #4 peak

    basic board with future potential built in - you can buy Arduino shields to do a wide variety of things - it doesn't have an Operating System (OS) per se. Arduino has a great deal of support for accessing the board and driving serv

  5. #5 peak

    28 "And you shall say to them, 'This is the nation that did not obey the voice of the LORD their God or accept correction; truth has perished and has been cut

  6. #6 peak

    ic Buddhist Apologetics (PDF)⏎Antinomianism comes in at least two forms: strong or weak. Weak antinomianism is permissive: one may do what is bad and nevertheless somehow remain not guilty. Strong

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: how to make a single single string, but not through a single string, but through a single string, but through a single string, but through a single string, but through a single string, but through a single string, but through 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

    iking Trails for Dogs||5 Ways to Get Your Cat with Mobility Issues Moving||5 Ways to Get Your Dog with Mobility Issues Moving||5 Exercise Tips for Arthritic Cats|⏎|Top Ten 'Small Breed' Dogs||Top Ten Dogs for

  2. #2 peak

    990 to 2019. The researchers analyzed life expectancy data for the 3 largest race/ethnicity groups by state: Hispanic, non-Hispanic Black, and non-Hispanic White Americans. They found that although

  3. #3 peak

    face of a threat. Birds raised in an urban environment react differently than country-bred birds when faced with a predator.⏎A study undertaken by Diego Ibanez-Alamo, researcher at the University of Granada (UGR) and Anders P

  4. #4 peak

    Dionysius II 65: οὔτε γὰπ τὸ χωρίον τοῦτο, ἐν ᾧ τὸ ἰερὸν φυ

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. L22:530 suppresses
  2. L21:1934 suppresses
  3. L17:257 suppresses
  4. L20:1120 excites
  5. L24:105 weakly suppresses
  6. L19:1351 weakly suppresses
  7. L13:363 weakly suppresses
  8. L7:2329 weakly excites
  9. L22:366 weakly excites
  10. L7:1982 weakly suppresses
  11. L16:2974 weakly excites
  12. L8:2295 weakly excites
  13. L18:811 weakly excites
  14. L21:237 weakly suppresses
  15. L14:178 weakly suppresses
  16. L21:2715 weakly suppresses
  17. L21:2323 weakly excites
  18. L2:265 weakly suppresses
  19. L16:1754 weakly excites
  20. L13:2346 weakly suppresses
  21. L14:373 weakly suppresses
  22. L18:986 weakly excites
  23. L19:2348 weakly excites
  24. L15:12 weakly excites

Feeds later layers

  1. L27:2799 weakly suppresses the letter d inside words
  2. L26:3066 weakly excites
  3. L26:294 weakly excites
  4. L26:2936 weakly excites
  5. L27:915 weakly suppresses unclear: mixed word fragments and phrases
  6. L27:3014 weakly suppresses
  7. L26:2980 weakly suppresses
  8. L27:2308 weakly suppresses
  9. L27:1074 weakly suppresses
  10. L27:919 weakly suppresses
  11. L27:1515 weakly excites
  12. L26:1523 weakly excites
  13. L27:526 weakly excites
  14. L27:1426 weakly suppresses
  15. L27:41 weakly suppresses
  16. L27:501 weakly suppresses
  17. L26:2392 weakly suppresses
  18. L27:1105 weakly suppresses
  19. L26:2338 weakly suppresses
  20. L26:466 weakly excites
  21. L27:263 weakly suppresses
  22. L26:930 weakly excites
  23. L26:1688 weakly excites
  24. L27:146 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 -9.50 and -8.761 values between -5.79 and -5.054 values between -5.05 and -4.311 values between -4.31 and -3.573 values between -3.57 and -2.823 values between -2.82 and -2.084 values between -2.08 and -1.3480 values between -1.34 and -0.603049 values between -0.60 and 0.14724 values between 0.14 and 0.8921 values between 0.89 and 1.638 values between 1.63 and 2.372 values between 2.37 and 3.111 values between 4.59 and 5.341 values between 6.82 and 7.561 values between 7.56 and 8.301 values between 8.30 and 9.052 values between 9.05 and 9.791 values between 19.43 and 20.17offnegativepositiveoften

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. 真的
  2. лаг
  3. ля
  4. 真
  5. ไหล
  6. carg
  7. ham
  8. дав
  9. grátis
  10. htar

Suppresses

  1. alytics
  2. isations
  3. ��
  4. .appspot
  5. 消
  6. /array
  7. resenter
  8. eding
  9. /course
  10. conde