Neuron Explorer

Qwen3-0.6B

capital letters inside acronyms

A neuron in layer 26 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 responds to capital letters in the middle of a word or abbreviation, such as the 'H' in HSC, the 'OR' in ORBIMAGE, or the 'RO' in ROOSEVELT. Turning it up did not produce a clear effect; the text mostly repeated words or drifted.

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

    asked to critically think and debate varying interpretations of the empirical evidence in relation to the science being brought to your attention<|endoftext|>Economics– Important Questions Bank for Rajasthan Board HSC 201

  2. #2 peak

    what constitutes fair use of a copyrighted work involves examining and balancing the effects of four different relevant factors:⏎Retrieved and used with permission from ASCD http://www.ascd.org/publications/books/1

  3. #3 peak

    almost all the schools of philosophy. This dogma has been described in the Vyāsabhāsya on the Yogasutras of Patañjali (200 B. C.). The word ‘ekabhavika

  4. #4 peak

    ZOOMER:<|endoftext|>SeaWiFS: Gulf Coast Sediments⏎Provided by the SeaWiFS Project, NASA/Goddard Space Flight Center, and ORBIMAGE⏎Various shades of brown color the water along the Louisiana/Mississippi

  5. #5 peak

    20system.⏎- How Vitamin C Supports a Healthy Immune Sysstem. Academy of Nutrition and Dietetics. 2020 https://www.eatright.org/food/vitamins-and-supplements/types-of-vitamins-and

  6. #6 peak

    to protect the interests of the people, and a people strong enough and well enough informed to maintain its sovereign control over its government.⏎President FRANKLIN D. ROOSEVELT, fireside chat on economic conditions, April 14,

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: cut back cut back cut back cut back cut back cut back cut back cut back cut back cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut cut

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

    nder.⏎- Buzzle --> Klebsi<|endoftext|>WHAT IS SOLAR PV?⏎Solar PV panels convert light into electricity.⏎WHAT IS THE DIFFERENCE BETWEEN SOLAR PV AND SOLAR THERMAL.⏎Solar PV panels use daylight to generate electricity.

  2. #2 peak

    amount of corrosive elements.The material will be placed underwater or be exposed to water consistently.In applications where greater strength and hardness are required.The Difference Between 304 and 316 Stainless Steel Metal results for this questionWhat is electrical conductivity

  3. #3 peak

    a typhoon if it happened in the northwestern Pacific.⏎WHAT’S THE DIFFERENCE BETWEEN A HURRICANE AND A TROPICAL STORM?⏎Wind speed. When tropical cyclones are

  4. #4 peak

    s PART THREE: RESEARCH DESIGN The Qualitative Research Process Research Questions Entering the Field Sampling How to Design Qualitative Research: An Overview PART FOUR: VERBAL DATA Interviews Narratives Focus Groups Verbal Data: An Overview PART FIVE:

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. L23:509 strongly suppresses
  2. L22:3045 suppresses
  3. L20:473 suppresses
  4. L22:376 excites
  5. L24:353 excites decimal point inside a number
  6. L19:1753 suppresses
  7. L22:67 suppresses
  8. L19:68 excites full stops closing a claim
  9. L19:2069 excites
  10. L21:605 suppresses
  11. L25:6 suppresses
  12. L24:421 suppresses
  13. L19:1930 suppresses
  14. L23:179 excites
  15. L20:2221 suppresses
  16. L24:192 suppresses
  17. L19:2730 suppresses
  18. L22:430 suppresses
  19. L23:2137 suppresses
  20. L25:1270 suppresses
  21. L9:11 suppresses
  22. L24:2168 excites
  23. L18:946 suppresses
  24. L24:1403 weakly excites

Feeds later layers

  1. L27:1369 strongly suppresses
  2. L27:1448 suppresses
  3. L27:1956 suppresses
  4. L27:2212 weakly suppresses
  5. L27:2733 weakly suppresses
  6. L27:899 weakly suppresses
  7. L27:643 weakly suppresses
  8. L27:568 weakly suppresses
  9. L27:877 weakly excites
  10. L27:439 weakly suppresses
  11. L27:1768 weakly suppresses
  12. L27:2339 weakly excites
  13. L27:94 weakly suppresses
  14. L27:1090 weakly suppresses
  15. L27:2357 weakly suppresses
  16. L27:2241 weakly suppresses
  17. L27:1970 weakly suppresses
  18. L27:1779 weakly suppresses
  19. L27:1055 weakly suppresses
  20. L27:851 weakly suppresses first names of famous people
  21. L27:1872 weakly suppresses
  22. L27:1565 weakly suppresses
  23. L27:745 weakly excites
  24. L27:856 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 -19.13 and -18.251 values between -17.38 and -16.505 values between -15.63 and -14.757 values between -14.75 and -13.885 values between -13.88 and -13.009 values between -13.00 and -12.136 values between -12.13 and -11.255 values between -11.25 and -10.3813 values between -10.38 and -9.5013 values between -9.50 and -8.6314 values between -8.63 and -7.7529 values between -7.75 and -6.8825 values between -6.88 and -6.0024 values between -6.00 and -5.1335 values between -5.13 and -4.2534 values between -4.25 and -3.3862 values between -3.38 and -2.5087 values between -2.50 and -1.63278 values between -1.63 and -0.752281 values between -0.75 and 0.12864 values between 0.12 and 1.0055 values between 1.00 and 1.8723 values between 1.87 and 2.759 values between 2.75 and 3.627 values between 3.62 and 4.503 values between 4.50 and 5.374 values between 5.37 and 6.243 values between 7.99 and 8.871 values between 8.87 and 9.741 values between 11.49 and 12.372 values between 12.37 and 13.241 values between 14.12 and 14.991 values between 14.99 and 15.87offnegativepositiveoften

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. izin
  2. oin
  3. userManager
  4. 化进程
  5. ihan
  6. に入
  7. aghan
  8. iban
  9. okit
  10. dictionaries

Suppresses

  1. And
  2. And
  3. -and
  4. _and
  5. .And
  6. _And
  7. _AND
  8. AND
  9. But
  10. Or