Neuron Explorer

Qwen3-0.6B

'other' after a preposition

A neuron in layer 2 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 word "other" when it follows a word like "to", "from" or "of", as in "transferred to other flowers". When it was turned up, the written text did not clearly move toward that topic, so the link to the steered writing is unclear.

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

    petals of the flower are designed to attract insects and bees, so that pollen can be transferred to other flowers. This is called pollination.⏎Pollination is needed for plants to reproduce

  2. #2 peak

    of freely mixing bats across the entire continent facilitates the spread of the viruses. This has important implications for the monitoring of these viruses in order to prevent its spread to other animals, including humans.”⏎Fruit bats are often hunted for meat, a process

  3. #3 peak

    without which we cannot survive.⏎So we need to learn about our origins⏎By understanding our shared roots in Africa, and by learning about our more ancient connections to other branches on the tree of life, we can find unity among ourselves, and resolve

  4. #4 peak

    that is by all accounts assuming control over, bit by bit, all enterprises. Because of its numerous points of interest, plastic regularly gives better options in contrast to other packaging materials, and it is utilized for apparently unlimited applications. There are a few

  5. #5 peak

    never boring also thanks to its wide variety of subjects. Some astronomers study volcanoes on Mars, others examine stars, while there are also astronomers examining the existence of other universes. During their researches, some scientists have, accidentally or not,

  6. #6 peak

    fatty acids can help to minimize symptoms.⏎Why are so many alcoholics skinny?⏎More likely, it results from interference with the body’s ability to derive energy from other foods. According to Lieber’s report, experiments in laboratory animals and in heavy

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: and 1 to sort of. 1. . to sort. . . . . . ., and 1. ... . . . . ., and 1. . . . . . ., and 1. .

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

    .⏎Metsäpohjainen piha paransi päiväkotilasten immuunijärjestelmää

  2. #2 peak

    Maladie pour 2014 – Rapport au Ministre chargé de la Sécurité Sociale et au Parlement sur l’évolution des charges et produits de l’Assurance Maladie au titre de 2

  3. #3 peak

    remuage phase.⏎A research carried out by the Department of Agricultural and Environmental Sciences - Production, Landscape, Agroenergy of the Università degli Studi di Milano aims at providing an innovative remuage method which is alternative

  4. #4 peak

    an information processing system that PERCEIVES, THINKS, REASONS AND DECIDES OR FORMS CONCLUSIONS.⏎Thus, when a bird builds a birds' nest it does so by processing information with its mind. The sticks and tw

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. L1:26 suppresses
  2. L1:805 excites
  3. L1:2508 suppresses
  4. L0:82 suppresses
  5. L0:733 suppresses
  6. L1:518 excites
  7. L1:2860 excites
  8. L1:1372 excites
  9. L1:261 excites
  10. L1:1794 excites
  11. L0:1097 suppresses
  12. L1:1665 suppresses
  13. L1:683 excites
  14. L1:2624 weakly excites
  15. L0:1831 weakly suppresses
  16. L0:425 weakly excites
  17. L1:2607 weakly suppresses
  18. L1:1274 weakly excites
  19. L1:740 weakly excites
  20. L1:1686 weakly suppresses
  21. L0:2391 weakly suppresses
  22. L1:2194 weakly excites
  23. L0:1910 weakly excites
  24. L0:954 weakly excites

Feeds later layers

  1. L4:298 strongly excites
  2. L3:19 strongly excites
  3. L5:13 strongly excites
  4. L9:55 strongly excites
  5. L10:661 strongly excites
  6. L8:253 strongly suppresses
  7. L17:2744 strongly excites 'of' after 'those' or 'these'
  8. L12:1587 suppresses
  9. L10:246 excites
  10. L11:282 excites
  11. L11:1684 excites
  12. L4:450 suppresses
  13. L6:424 excites
  14. L10:1190 suppresses
  15. L6:450 suppresses
  16. L18:1096 excites
  17. L18:445 excites
  18. L5:194 suppresses 'a' or 'of' right after a verb
  19. L17:147 excites
  20. L5:105 suppresses
  21. L16:341 excites
  22. L16:1418 suppresses
  23. L9:2445 excites
  24. L19:1629 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 -1.77 and -1.691 values between -0.98 and -0.911 values between -0.83 and -0.762 values between -0.76 and -0.692 values between -0.62 and -0.554 values between -0.55 and -0.484 values between -0.48 and -0.4112 values between -0.41 and -0.3332 values between -0.33 and -0.2685 values between -0.26 and -0.19272 values between -0.19 and -0.12601 values between -0.12 and -0.051587 values between -0.05 and 0.03827 values between 0.03 and 0.10294 values between 0.10 and 0.17106 values between 0.17 and 0.2435 values between 0.24 and 0.3115 values between 0.31 and 0.386 values between 0.38 and 0.4610 values between 0.46 and 0.534 values between 0.53 and 0.603 values between 0.60 and 0.671 values between 0.67 and 0.741 values between 0.89 and 0.962 values between 1.03 and 1.10offnegativepositiveoften

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. 比如
  2. aren
  3. pecies
  4. 也都
  5. ibling
  6. personalities
  7. ');</
  8. 嘉宾
  9. oods
  10. criptions

Suppresses

  1. itself
  2. ус
  3. 日报道
  4. signed
  5. 一名
  6. general
  7. aviest
  8. sí
  9. овая
  10. 期