Neuron Explorer

Qwen3-0.6B

end of citations and titles

A neuron in layer 20 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron fires at the point where a book title, file name or reference ends, such as a full stop, comma or dash just after a title. It is a loose pattern, and the turned-up steering output did not show a clear link to it.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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

    desires, the right kind method to pursue happiness, and the correct plan for an excellent lifestyles.⏎Read or Download Aristotle for Everybody: Difficult Thought Made Easy PDF⏎Best greek &

  2. #2 peak

    File Size: KB.⏎Get this from a library! Principles of biblical interpretation: (sacred hermeneutics). [Louis Berkhof] -- This is a textbook in Biblical interpretation. It is specifically designed for class use in

  3. #3 peak

    OLAP queries - Metadata repository – DWH Implementation - OLAM⏎1. JiaweiHan ,MichelineKamber, "Data Mining: Concepts and Techniques", 2nd Edition, Elsevier India Private Limited,2008.⏎

  4. #4 peak

    ? It is up to me and you.”⏎Adafruit publishes a wide range of writing and video content, including interviews and reporting on the maker market and the wider technology world. Our standards page is intended as a guide to best practices that Adafruit

  5. #5 peak

    smooth variations which are precise to the unique paintings.⏎Sextus Empiricus' Outlines of Pyrrhonism is among the most crucial and influential texts within the historical past of Greek philosophy. within the calls for of cause Casey Perin

  6. #6 peak

    had little effect but to strengthen comparable hard-line views in the Kremlin.⏎Armstrong, David, and Erik Goldstein, eds. The End of the Cold War. Routledge, 2013.⏎

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: "The answer to the question "What is the answer of the series "This Is Our Sunday" of the series "This Is a series of series?" a series of series?" **** **** **** **** **** ****

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

    For further information and to get help call the Poison Help Line at 1-800-222-1222.⏎Syn

  2. #2 peak

    mass production. The first two industrial revolutions made people richer and more urban. Now a third revolution is under way. Manufacturing is going digital. As this week’s special report argues, this could change not just business, but much else besides.⏎A

  3. #3 peak

    in cancer treatment and outcomes due to race, ethnicity, socioeconomic status, sexual orientation and more. The report also includes a special section on COVID-19.⏎On the steering committee for this important report was

  4. #4 peak

    6-3008; doi:10.3390/su5073006⏎I am honored to introduce this special issue of Sustainability, which exemplifies how the field of Psychology can contribute

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. L18:2611 suppresses
  2. L18:982 excites
  3. L4:1050 excites
  4. L17:591 weakly suppresses
  5. L3:1215 weakly excites
  6. L15:2437 weakly suppresses opening parenthesis after a noun
  7. L7:2259 weakly suppresses
  8. L0:3036 weakly excites
  9. L5:2406 weakly suppresses
  10. L18:913 weakly suppresses
  11. L8:84 weakly suppresses
  12. L1:1530 weakly suppresses
  13. L2:2462 weakly suppresses
  14. L15:2445 weakly suppresses
  15. L15:682 weakly excites
  16. L14:2645 weakly excites
  17. L6:2545 weakly suppresses
  18. L11:51 weakly suppresses
  19. L10:2550 weakly suppresses
  20. L5:2273 weakly excites
  21. L2:94 weakly suppresses
  22. L7:1807 weakly excites
  23. L4:1188 weakly excites
  24. L17:1683 weakly suppresses

Feeds later layers

  1. L24:2621 weakly suppresses
  2. L27:1349 weakly suppresses
  3. L22:1125 weakly suppresses
  4. L26:382 weakly suppresses
  5. L27:1041 weakly suppresses
  6. L22:1565 weakly suppresses
  7. L23:1816 weakly suppresses
  8. L22:757 weakly suppresses
  9. L21:2746 weakly suppresses
  10. L27:2241 weakly suppresses
  11. L23:650 weakly suppresses
  12. L22:2080 weakly suppresses
  13. L26:1106 weakly suppresses
  14. L22:796 weakly excites
  15. L24:1832 weakly suppresses unclear: mixed multilingual, mid-word splits
  16. L27:824 weakly suppresses
  17. L27:2397 weakly suppresses
  18. L22:1998 weakly suppresses
  19. L24:1100 weakly suppresses
  20. L21:1341 weakly suppresses
  21. L25:1107 weakly suppresses
  22. L23:2442 weakly suppresses
  23. L25:2537 weakly suppresses
  24. L22:2033 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 -22.19 and -21.401 values between -17.45 and -16.661 values between -15.87 and -15.082 values between -15.08 and -14.301 values between -14.30 and -13.512 values between -13.51 and -12.721 values between -12.72 and -11.932 values between -11.93 and -11.144 values between -11.14 and -10.354 values between -10.35 and -9.564 values between -9.56 and -8.775 values between -8.77 and -7.982 values between -7.98 and -7.196 values between -7.19 and -6.4011 values between -6.40 and -5.6111 values between -5.61 and -4.8221 values between -4.82 and -4.0424 values between -4.04 and -3.2539 values between -3.25 and -2.4663 values between -2.46 and -1.67125 values between -1.67 and -0.881576 values between -0.88 and -0.091867 values between -0.09 and 0.7092 values between 0.70 and 1.4926 values between 1.49 and 2.2812 values between 2.28 and 3.074 values between 3.07 and 3.861 values between 8.59 and 9.38offnegativepositiveoften

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. ivamente
  2. OU
  3. óm
  4. ainen
  5. �单
  6. ạn
  7. alguém
  8. AML
  9. 谈起
  10. ataka

Suppresses

  1. idd
  2. 答题
  3. answering
  4. vide
  5. _FEATURE
  6. /features
  7. uide
  8. 专栏
  9. /editor
  10. imedia