Neuron Explorer

Qwen3-0.6B

'in' after a named author or speaker

A neuron in layer 6 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 "in" right after a person's name or a statement verb, as in "wrote in" or "said in", where a source or publication is about to be named. Turning it up did not produce a clear visible change in the stories.

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

    have properties that increase diabetes risk independently from obesity.⏎The study was the icing on the cake for theories that sugar is toxic. As columnist Mark Bittman wrote in The New York Times:⏎The study demonstrates [that sugar causes diabetes] with the

  2. #2 peak

    JUNE 22, 1970, President Nixon extended the Voting Rights Act to let 18-year-olds vote.⏎The Supreme Court, in Oregon v Mitchell, limited this right so the 26th Amendment was passed

  3. #3 peak

    although Cox asserts that he and the physicists at CERN will prove or disprove their existence by 2017. However, as Anthony Kaufman writes in his review on the

  4. #4 peak

    if you have fibromyalgia.⏎Dr. Jim Sublett, chief of allergy immunology at the University of Louisville School of Medicine in Kentucky has repeatedly said in literature that over-the-counter remedies can have adverse side effects.⏎The side effects

  5. #5 peak

    every nation on the planet, and demonstrated Earth’s gravity need not prevent humanity from exploring the entire universe, US-based Space Foundation CEO Tom Zelibor had said in April.⏎On April 12, 1961, Gagar

  6. #6 peak

    riveting story of the two crusading lawyers who led the legal battle to end segregation, one case and one courtroom at a time.⏎The Supreme Court’s decision inBrown v. Board of Educationis widely considered a seminal point in the battle to

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: it's a function of the number of times I have been given the same amount of money. So, I have been given the function of the number of times I have been given the same amount of money. This is a function of 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

    permitting it for research done on existing stem cell lines, that is, stem cell lines created from embryos that have already been destroyed. The following is reprinted with permission from The Chicago Jewish News.⏎Pikuach nefesh. Preservation of life.

  2. #2 peak

    the overarching story formed from controlling stories within a culture. He argues that metanarrative is the basis for worldview core assumptions about all of reality. Matthews borrows from Michael Kearney who understands worldview to be the encompassing assumptions about: self

  3. #3 peak

    Hydrological Society⏎This article has been published in Journal of Hydrology (NZ). © 1995 The New Zealand Hydrological Society. Used with permission.<|endoftext|>A group at Lawrence Livermore National Laboratory that predicts the spread of hazardous

  4. #4 peak

    in children: time trend analysis. BMJ. August 6, 2005;331:328–9.⏎Used with permission from

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. L5:2358 suppresses
  2. L5:898 suppresses
  3. L5:353 weakly suppresses
  4. L5:26 weakly excites
  5. L5:2261 weakly excites word pieces inside medical and science terms
  6. L2:2668 weakly excites
  7. L5:1512 weakly excites
  8. L5:141 weakly excites
  9. L3:407 weakly excites
  10. L2:2414 weakly suppresses
  11. L1:1353 weakly excites
  12. L4:560 weakly suppresses
  13. L3:1488 weakly suppresses
  14. L0:2420 weakly suppresses
  15. L1:2180 weakly excites
  16. L0:2414 weakly excites
  17. L4:1035 weakly excites 'for' after verbs like 'fall' or 'click'
  18. L2:2503 weakly excites
  19. L0:372 weakly suppresses
  20. L4:1153 weakly suppresses
  21. L0:2233 weakly suppresses
  22. L0:951 weakly suppresses
  23. L5:1323 weakly excites
  24. L3:125 weakly excites verb right after a subject in clauses

Feeds later layers

  1. L7:2571 excites
  2. L7:2492 weakly excites
  3. L9:1873 weakly suppresses
  4. L12:1288 weakly excites
  5. L8:886 weakly excites
  6. L11:1699 weakly excites
  7. L11:1765 weakly excites
  8. L8:675 weakly excites
  9. L10:1832 weakly suppresses
  10. L12:2632 weakly excites
  11. L7:2369 weakly excites comma or 'to' before a quotation
  12. L12:920 weakly excites
  13. L12:2088 weakly excites
  14. L10:935 weakly suppresses
  15. L7:528 weakly excites
  16. L13:1251 weakly excites
  17. L12:1086 weakly excites
  18. L10:2770 weakly suppresses
  19. L15:1963 weakly excites
  20. L13:1349 weakly suppresses
  21. L11:1231 weakly excites
  22. L20:2273 weakly excites
  23. L7:198 weakly suppresses
  24. L7:1416 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 -1.70 and -1.572 values between -1.30 and -1.173 values between -1.17 and -1.043 values between -1.04 and -0.914 values between -0.91 and -0.775 values between -0.77 and -0.649 values between -0.64 and -0.5134 values between -0.51 and -0.38107 values between -0.38 and -0.25442 values between -0.25 and -0.111650 values between -0.11 and 0.021126 values between 0.02 and 0.15329 values between 0.15 and 0.2895 values between 0.28 and 0.4124 values between 0.41 and 0.5520 values between 0.55 and 0.6816 values between 0.68 and 0.8112 values between 0.81 and 0.944 values between 0.94 and 1.074 values between 1.07 and 1.211 values between 1.21 and 1.343 values between 1.34 and 1.471 values between 1.47 and 1.602 values between 1.60 and 1.734 values between 1.73 and 1.871 values between 2.39 and 2.532 values between 2.53 and 2.662 values between 2.66 and 2.791 values between 2.92 and 3.051 values between 3.45 and 3.58offnegativepositiveoften

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. directives
  2. 文中
  3. 窑
  4. asco
  5. yet
  6. .onNext
  7. 深度
  8. Cups
  9. 倾
  10. back

Suppresses

  1. имя
  2. idental
  3. _);⏎
  4. etchup
  5. _); ⏎
  6. oteca
  7. ENCH
  8. chooser
  9. ocha
  10. uspend