Neuron Explorer

Qwen3-0.6B

'with' after bring/carry/take

A neuron in layer 10 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 fires on the word 'with' when it follows verbs like 'bring', 'carry' or 'take', as in 'brought with them'. Turning it up made the model produce odd, self-referential text about 'itself' and 'myself', which fits its output words.

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

    cool red-hot robots. They discover the science of circuits and how robots use sensors to explore things around them. They build their very own working robot to take home with them! For kids entering 2nd-6th grade in fall 2

  2. #2 peak

    a style of therapy that helps people become—and embrace being—an expert in their own lives In narrative therapy, there is an emphasis on the stories we develop and carry with us through our lives⏎What theory is narrative therapy based on?⏎First developed by

  3. #3 peak

    use their leisure time and at the same time pursue health and fitness interests. Swimming interests Americans of all classes, ages, races and social status. Swimming has brought with it competitions that bring awards to America and to many individuals, swimming is a full

  4. #4 peak

    of Biology and Director of Sustainability Studies Rich Niesenbaum manages plants at the College's new pollinator habitat.⏎Summer has arrived in the Lehigh Valley, bringing with it sunshine, blooming flowers and buzzing insects. Rich Niesenbaum,

  5. #5 peak

    . The early settlers of Chortitza and Molotschna brought with them a considerable number of East Frisian cattle. Under the leadership of Corn

  6. #6 peak

    of white rays is dispersed by a prism which colour will be refracted to a larger extent?Answer: Violet.⏎This what would happen if humans disappeared essay carries with it immense complications owing to the Suns what would happen if humans disappeared essay on the

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: (10) (10) 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,

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

    bright, mocking beauty.⏎October 2003 | Volume 54, Issue 5⏎Your example of occupy brings to mind the way in which new words are formed. For example, the word ro

  2. #2 peak

    that cover the same entity or activity. The business register is of vital importance in such linking of microdata.⏎13.18. Compilers should take into account all relevant dimensions of data sources before merging; for example, the content of

  3. #3 peak

    Released at 11:30 AM (CANBERRA TIME) 24/04/2001⏎|Page tools: Print Page RSS Search this Product|⏎This variable identifies holders of Australian citizenship.⏎Citizens

  4. #4 peak

    publications it appears that Chloroquine prevents the gain of the virus into the cells and, therefore, prevents the cellular infection from occurring. We also have to take into consideration that the use of this medication has to be

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. L7:1090 strongly excites
  2. L4:1826 excites
  3. L8:1173 excites
  4. L4:1022 suppresses
  5. L2:1586 weakly suppresses
  6. L5:2354 weakly excites
  7. L3:1967 weakly suppresses
  8. L5:2338 weakly suppresses
  9. L9:2267 weakly suppresses 'with' after exchange nouns
  10. L9:1757 weakly suppresses
  11. L4:660 weakly suppresses
  12. L1:808 weakly suppresses
  13. L6:1482 weakly excites
  14. L4:16 weakly excites
  15. L8:14 weakly excites
  16. L4:2719 weakly excites
  17. L8:2045 weakly excites
  18. L7:2763 weakly excites
  19. L6:13 weakly excites
  20. L2:1450 weakly suppresses
  21. L6:1895 weakly excites
  22. L5:258 weakly excites
  23. L5:616 weakly excites
  24. L9:929 weakly excites 'each' in 'each other'

Feeds later layers

  1. L22:643 excites
  2. L21:1144 excites
  3. L11:173 excites
  4. L16:592 excites
  5. L14:677 suppresses
  6. L12:887 excites
  7. L27:1381 excites
  8. L17:259 excites
  9. L20:1476 excites
  10. L16:415 weakly excites
  11. L18:2290 weakly excites
  12. L25:849 weakly excites
  13. L19:541 weakly excites
  14. L23:1062 weakly excites
  15. L23:2992 weakly suppresses unclear: first pieces of split words
  16. L18:513 weakly excites
  17. L13:1513 weakly excites
  18. L18:1026 weakly excites
  19. L25:2825 weakly suppresses
  20. L14:611 weakly excites
  21. L27:631 weakly suppresses
  22. L15:364 weakly excites
  23. L20:272 weakly excites
  24. L14:248 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 -2.77 and -2.672 values between -0.99 and -0.893 values between -0.89 and -0.799 values between -0.79 and -0.6924 values between -0.69 and -0.5953 values between -0.59 and -0.49116 values between -0.49 and -0.40236 values between -0.40 and -0.30462 values between -0.30 and -0.20771 values between -0.20 and -0.101177 values between -0.10 and 0.00727 values between 0.00 and 0.10230 values between 0.10 and 0.2052 values between 0.20 and 0.3023 values between 0.30 and 0.4011 values between 0.40 and 0.502 values between 0.50 and 0.602 values between 0.60 and 0.692 values between 0.69 and 0.791 values between 0.79 and 0.891 values between 0.89 and 0.992 values between 0.99 and 1.091 values between 1.09 and 1.19offnegativepositiveoften

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. itself
  2. himself
  3. 自己
  4. oneself
  5. myself
  6. self
  7. 내
  8. sich
  9. herself
  10. собой

Suppresses

  1. мент
  2. rowsers
  3. LookAndFeel
  4. 第一家
  5. CollectionView
  6. PackageManager
  7. pez
  8. toLocale
  9. 玘
  10. arshal