Neuron Explorer

Qwen3-0.6B

colon in a book or article title

A neuron in layer 3 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 a colon that separates the main title of a book, article or report from its subtitle, such as "Machine Made: Tammany Hall...". Turning it up did not produce a clear change in the model's replies.

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

    politicians who kept that famously well-oiled patronage machine of old working for decades at least did something to earn their graft.⏎In a new book, Machine Made: Tammany Hall and the Creation of Modern American Politics, (Norton, New

  2. #2 peak

    3th USENIX Security Symposium, 2004.⏎Garfinkel, S. and Shelat, A., "Remembrance of Data Passed: A Study of Disk Sanitization Practices," IEEE Security and Privacy, January/

  3. #3 peak

    itamins-and-nutrients/how-vitamin-c-supports-a-healthy-immune-system⏎- Kennedy, David O. “B Vitamins and the Brain: Mechanisms, Dose and Efficacy—A Review.” Nutrients,

  4. #4 peak

    Life.com and the new book Plastic-Free: How I Kicked the Plastic Habit and How You Can Too (Skyhorse).

  5. #5 peak

    on the book "Dots and Dashes: Adventures at the Dawn of the Wireless Age" is at http://vubooks

  6. #6 peak

    and editorial matter, Kerry J. Kennedy and John Chi-Kin Lee; individual chapters, the contributors.⏎|Title of host publication||Religious education in Asia: Spiritual diversity in global

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 not. I need to get a better answer. I need to get a better answer. I need to get a better answer. I need to get a better answer. # I need to get a better answer. Here's 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

    Philadelphia, Pa: Elsevier Saunders; 2010:chap 73.⏎Karcher DS. McPherson<|endoftext|>Discrimination and disparity within correctional institutions

  2. #2 peak

    niere, at the River May (now St. John’s, Florida), in 1564. […]⏎Jacques le Moyne de Morgues (c. 1533–1588) was a

  3. #3 peak

    symptom relief, with the least restrictive diet possible.<|endoftext|>Evaluating Tagging Methods and Movement Patterns of Round Gobies⏎The round goby (Neogobius melanostomus) is an invasive fish species introduced in the St

  4. #4 peak

    Until well into the 20th century, knowledge of Jacques Le Moyne de Morgues was extremely limited,<|endoftext|>THE FATEFUL DECADE⏎THE FATEFUL

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. L2:111 suppresses
  2. L2:75 weakly suppresses
  3. L0:769 weakly suppresses
  4. L0:491 weakly suppresses
  5. L1:988 weakly excites
  6. L1:1957 weakly excites
  7. L2:1321 weakly suppresses
  8. L0:1224 weakly suppresses
  9. L1:2128 weakly excites
  10. L1:1279 weakly excites
  11. L2:852 weakly excites
  12. L0:1301 weakly suppresses
  13. L2:483 weakly suppresses
  14. L2:560 weakly excites
  15. L2:430 weakly excites
  16. L0:1197 weakly suppresses
  17. L0:178 weakly suppresses
  18. L0:36 weakly suppresses
  19. L2:70 weakly excites
  20. L1:2270 weakly excites
  21. L2:998 weakly suppresses
  22. L0:869 weakly excites
  23. L2:1564 weakly excites
  24. L2:911 weakly excites

Feeds later layers

  1. L5:2629 weakly excites
  2. L13:1501 weakly suppresses
  3. L15:1594 weakly suppresses
  4. L10:2365 weakly excites
  5. L27:2422 weakly suppresses
  6. L25:1612 weakly suppresses
  7. L13:2036 weakly excites
  8. L27:1795 weakly suppresses
  9. L11:1721 weakly suppresses
  10. L26:1689 weakly suppresses
  11. L19:1117 weakly excites
  12. L4:2163 weakly suppresses
  13. L23:1565 weakly suppresses
  14. L15:1645 weakly suppresses
  15. L14:1921 weakly excites
  16. L12:440 weakly excites
  17. L12:228 weakly excites end-of-document token after web boilerplate
  18. L20:1387 weakly excites
  19. L23:2871 weakly suppresses
  20. L21:1892 weakly excites
  21. L16:2112 weakly suppresses
  22. L17:135 weakly excites
  23. L9:2882 weakly excites
  24. L8:1305 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 -1.25 and -1.181 values between -0.98 and -0.911 values between -0.84 and -0.777 values between -0.56 and -0.497 values between -0.49 and -0.4314 values between -0.43 and -0.3652 values between -0.36 and -0.29113 values between -0.29 and -0.22279 values between -0.22 and -0.15572 values between -0.15 and -0.081027 values between -0.08 and -0.011110 values between -0.01 and 0.06479 values between 0.06 and 0.13157 values between 0.13 and 0.1952 values between 0.19 and 0.2617 values between 0.26 and 0.335 values between 0.33 and 0.403 values between 0.40 and 0.472 values between 0.47 and 0.542 values between 0.54 and 0.611 values between 0.61 and 0.681 values between 0.88 and 0.951 values between 0.95 and 1.021 values between 1.02 and 1.091 values between 1.09 and 1.161 values between 1.36 and 1.431 values between 1.43 and 1.50offnegativepositiveoften

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. rgan
  2. 不死
  3. idge
  4. Cas
  5. Cas
  6. otherwise
  7. Case
  8. southwest
  9. ensitive
  10. either

Suppresses

  1. =DB
  2. DBC
  3. "],["
  4. getBlock
  5. cular
  6. ',{⏎
  7. '],⏎
  8. huyết
  9. 搬
  10. ],[