Neuron Explorer

Qwen3-0.6B

comma after 'in this blog post'

A neuron in layer 19 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

The neuron fires on the comma that ends an opening phrase like 'In this blog post,' or 'Below,'. It is pushed the other way by the words 'a' and 'an' before a noun. Turning it up made the model write about unrelated math and turning it down gave fragments about movies and reports.

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

    . These innovations hold the promise of transforming industries, revolutionizing the way we live and work, and addressing some of society’s most pressing challenges. In this blog, we’ll take a deep dive

  2. #2 peak

    heritage resources (including scientifically significant paleontological resources).⏎We are currently undergoing a renaissance in the discovery, research, and interpretation of fossils from Utah.⏎Below, click on the various districts (West Desert, Green River, Color Country, Canyon

  3. #3 peak

    your hair? The intricate anatomy of our hair strands comprises several distinct components, each playing a unique role in the appearance and health of our hair. In this article, we will explore the essential elements of hair anatomy: the cuticle, cortex,

  4. #4 peak

    Quality⏎Common Resources — Feb 7, 2018⏎The Facts on Fracking: A Podcast with Daniel Raimi⏎In this podcast, RFF’s Daniel Raimi gives a balanced and accessible view of oil and

  5. #5 peak

    range of scholarship about Caddo history and culture from anthropologists, historians, and other researchers.⏎In this extensive 2021 bibliography, you will find 386 pages of resources to learn more about Cadd

  6. #6 peak

    to its former glory.Built in 1902, the hotel boasts a stunning atrium, luxurious amenities, and a fascinating history.In this blog post, we’ll take a closer look at the West Baden Hotel, its history,

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: there are a *there* *a* *a* *a* *a* *a* *a *a *a *a *a *a *a *a *a *a *a *a *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

    Junction. (A historical marker was placed at that spot in 1955.) News of Vilas’s work quickly spread; American Forestry magazine had an article about it

  2. #2 peak

    This is called parallax, and I’ve explained this in an earlier post for a solar eclipse that had very similar circumstances as this Swift observation. I’ve written a general parallax description as well.⏎So the wiggly path of the asteroid

  3. #3 peak

    aping Chesapeake Bay Watershed - U.S. Fish & Wildlife Service.⏎Here are some resources to help gardeners select native plants. The PA DCNR site has a listing of nurseries.⏎There are 133 species and subspecies

  4. #4 peak

    see some military time starting all the way back in 1873.⏎45-70 VS 444 Marlin⏎We have dedicated a post to the comparison of 45-70 vs 444

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. L17:805 suppresses
  2. L17:2970 weakly excites
  3. L7:935 weakly excites
  4. L17:1139 weakly excites
  5. L18:2293 weakly suppresses
  6. L10:609 weakly excites
  7. L16:2714 weakly suppresses
  8. L15:2387 weakly excites
  9. L6:2099 weakly excites
  10. L16:3002 weakly excites
  11. L18:1999 weakly suppresses
  12. L10:2818 weakly suppresses
  13. L0:1253 weakly excites surname pieces split across tokens
  14. L11:2626 weakly suppresses
  15. L14:1847 weakly excites
  16. L15:1889 weakly excites
  17. L18:2631 weakly suppresses
  18. L16:2824 weakly excites 'in' after 'published' or 'appears'
  19. L17:2607 weakly excites
  20. L11:976 weakly excites
  21. L17:679 weakly suppresses
  22. L18:262 weakly excites
  23. L13:1970 weakly suppresses
  24. L15:2484 weakly excites

Feeds later layers

  1. L22:778 suppresses
  2. L24:1263 suppresses
  3. L25:45 excites
  4. L25:1586 suppresses
  5. L27:797 weakly suppresses
  6. L22:103 weakly excites
  7. L20:1707 weakly suppresses
  8. L20:2682 weakly suppresses
  9. L25:407 weakly excites
  10. L24:823 weakly suppresses
  11. L26:2817 weakly suppresses
  12. L24:970 weakly suppresses
  13. L24:413 weakly suppresses
  14. L22:811 weakly suppresses
  15. L20:2696 weakly suppresses
  16. L20:1727 weakly suppresses
  17. L25:544 weakly suppresses
  18. L24:2538 weakly suppresses
  19. L27:716 weakly excites
  20. L25:2475 weakly excites
  21. L20:2000 weakly suppresses "a" or "the" opening an object noun phrase
  22. L21:1505 weakly excites
  23. L21:1100 weakly suppresses
  24. L27:756 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 -5.58 and -5.231 values between -4.88 and -4.531 values between -4.53 and -4.173 values between -4.17 and -3.821 values between -3.82 and -3.474 values between -3.12 and -2.763 values between -2.76 and -2.415 values between -2.41 and -2.063 values between -2.06 and -1.7110 values between -1.71 and -1.3619 values between -1.36 and -1.0062 values between -1.00 and -0.65411 values between -0.65 and -0.301910 values between -0.30 and 0.05820 values between 0.05 and 0.41247 values between 0.41 and 0.76150 values between 0.76 and 1.1180 values between 1.11 and 1.4665 values between 1.46 and 1.8233 values between 1.82 and 2.1725 values between 2.17 and 2.5216 values between 2.52 and 2.8710 values between 2.87 and 3.226 values between 3.22 and 3.585 values between 3.58 and 3.934 values between 3.93 and 4.282 values between 4.28 and 4.631 values between 4.63 and 4.993 values between 4.99 and 5.342 values between 5.34 and 5.691 values between 5.69 and 6.042 values between 6.04 and 6.391 values between 6.39 and 6.751 values between 8.16 and 8.51offnegativepositiveoften

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. -fontawesome
  2. добро
  3. ENDING
  4. iales
  5. 想不到
  6. lưới
  7. fec
  8. ending
  9. winnings
  10. etermined

Suppresses

  1. 有一个
  2. 有个
  3. 有
  4. an
  5. 有一
  6. a
  7. 也有
  8. 专门
  9. had
  10. 都有