Neuron Explorer

Qwen3-0.6B

'from' after origin verbs like imported or brought

A neuron in layer 11 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

Fires on the word 'from' when it names where something came from, as in 'imported from Mexico' or 'brought in from Canada'. Turning it up made the writing wander toward unrelated topics like new products and lists, so the link to origins is not carried through.

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

    Hybrid Tea Roses—were unknown. Early gardeners would have relied on musk roses, damask roses, and others. Ancient in pedigree, some were imported from China and the Middle East.⏎What is an Old Rose? Experts say that any

  2. #2 peak

    species of of wolves in the United States: gray wolves (Canis lupus) and red wolves (Canis rufus). Gray wolves were brought in from Canada to repopulate Yellowstone Park. DNA studies have now determined that the only

  3. #3 peak

    the Norway Rat has the head less elongated, a. blunter muzzle, and smaller ears than the Black Rat. The former was introduced into England - probably from the East Indies - early in the eighteenth century, and has not only almost

  4. #4 peak

    , three deaths and hundreds of illnesses in Pennsylvania, Tennessee and Georgia earlier this year were tied to an outbreak of hepatitis A, which was traced to green onions imported from Mexico and served in a restaurant.⏎To ensure<|endoftext|>Webster's Revised Unab

  5. #5 peak

    Australia’s Taronga Zoo. Right now were listening to the sounds of her feeding Dingoes.⏎“A dingo is a wild dog that was brought originally to Australia from Asia about five thousand years ago. It was used as as a pet in Asia

  6. #6 peak

    apweed, Canada thistle, African rue and leafy spurge. As their names indicate, weeds have arrived from well outside New Mexico's borders.⏎"We are on a threshold," says Richard Lee

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: "Do these data sets all have a median, but all the numbers 4 - 5 - 3 - 2 - 2 - 2 - 2 - 2 - 2 - 2 - 2 -

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

    -mails from anyone you don’t know or you otherwise aren’t sure about.⏎4. Watch out for sudden pop-up windows asking for personal information or warning of a virus.⏎5. Use a mix of security tools and procedures.⏎6. Beware

  2. #2 peak

    EAR) of Vitamin A in their daily diet. This trend is particularly worrying, given the rising rates of food insecurity that arose as a consequence of the COVID-19 pandemic: the September 2020 Social Weather Stations (SWS

  3. #3 peak

    child's progress.⏎- Step 1: Observe whether at 12 months the baby pokes, points, stands, grasps small objects, copies sounds and actions, tries words, and understands simple commands. Your baby should be testing

  4. #4 peak

    to as partial sight.⏎Cerebral Palsy. A group of conditions resulting from damage to the central nervous system. Do not assume that a person with cerebral palsy also has mental retardation; the two do not necessarily or typically occur together

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. L10:2602 suppresses
  2. L10:501 excites
  3. L10:2770 suppresses
  4. L7:1840 excites
  5. L8:158 excites
  6. L10:1357 suppresses
  7. L10:549 suppresses
  8. L7:729 suppresses
  9. L8:220 suppresses
  10. L10:1199 excites
  11. L9:755 suppresses
  12. L9:1080 excites
  13. L7:2786 excites
  14. L10:2639 excites
  15. L8:628 excites
  16. L9:906 suppresses
  17. L10:408 excites
  18. L0:2455 weakly excites
  19. L7:585 weakly excites
  20. L10:1701 weakly excites
  21. L8:954 weakly suppresses vitamin D mentions
  22. L7:1326 weakly suppresses
  23. L9:193 weakly excites
  24. L9:2883 weakly suppresses

Feeds later layers

  1. L13:451 suppresses
  2. L13:2955 excites
  3. L12:845 suppresses
  4. L12:667 suppresses
  5. L14:1968 excites
  6. L19:2320 excites
  7. L13:574 excites
  8. L14:1021 excites
  9. L12:767 excites
  10. L16:1493 suppresses
  11. L13:2522 suppresses
  12. L13:1147 suppresses
  13. L13:2839 excites
  14. L12:1930 excites
  15. L12:279 excites
  16. L22:1112 excites
  17. L14:2830 excites
  18. L12:1650 excites
  19. L13:430 suppresses
  20. L12:2489 excites
  21. L17:550 suppresses
  22. L13:2173 excites
  23. L14:1450 excites
  24. L25:1709 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.

2 values between -1.35 and -1.213 values between -1.07 and -0.933 values between -0.93 and -0.7813 values between -0.78 and -0.6443 values between -0.64 and -0.5089 values between -0.50 and -0.36361 values between -0.36 and -0.22884 values between -0.22 and -0.071870 values between -0.07 and 0.07416 values between 0.07 and 0.21115 values between 0.21 and 0.3539 values between 0.35 and 0.4917 values between 0.49 and 0.6415 values between 0.64 and 0.7810 values between 0.78 and 0.925 values between 0.92 and 1.063 values between 1.06 and 1.205 values between 1.20 and 1.354 values between 1.35 and 1.491 values between 1.49 and 1.631 values between 1.63 and 1.774 values between 1.77 and 1.911 values between 2.20 and 2.341 values between 2.34 and 2.481 values between 3.19 and 3.331 values between 3.90 and 4.041 values between 4.19 and 4.33offnegativepositiveoften

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. 进来
  2. 而来
  3. 告诉她
  4. 引入
  5. 来的
  6. มาจาก
  7. 到现在
  8. 充实
  9. 手里
  10. infl

Suppresses

  1. остоя
  2. furt
  3. uno
  4. ultiple
  5. phasis
  6. 被执行人
  7. uat
  8. ⋮
  9. {
  10. .setTag