Neuron Explorer

Qwen3-0.6B

unclear: weak, scattered triggers

A neuron in layer 2 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron responds only weakly and to a mix of unrelated words, such as 'subtract', 'parent', 'health' and 'aerodynamics'. No single pattern stands out, and turning it up or down gave no clear change in the text.

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

    0.⏎- Multiply it by 9. [Pause while they do this]⏎- If it's a two-digit number, add them together.⏎- Now, subtract 5 from the number in your head. [Pause again]⏎- Now,

  2. #2 peak

    away your identity.⏎- Tell your parent or another adult you trust if you see upsetting language, nasty pictures or something scary on the internet.⏎- Always ask a parent or adult before you fill out any forms, or give out money or credit card

  3. #3 peak

    same denominators. So click and give it a try. It's easier than you think.⏎Fractions - Subtracting Mixed Numbers⏎Do you wonder how to subtract mixed numbers? Do you know how to handle unlike denominators or how to reg

  4. #4 peak

    an electric car that will break – no, shatter – the existing E-1 class electric land speed record of 204 mph. By combining aerodynamics with new, cutting-edge green energy technology, the Electric Land Speed Record Car will

  5. #5 peak

    continue to grow throughout seven years of follow-up,” said study author Wendy King, an associate professor of epidemiology at the University of Pittsburgh’s Graduate School of Public Health.⏎The researchers say that because alcohol abuse cases don’t become apparent until many years

  6. #6 peak

    Canada; however, imported cases continue to occur.⏎To obtain or inquire about an immunization record, or for general immunization information, please contact your local public health unit.<|endoftext|>states that England officially recognizes the colonies to be in rebellion and that

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: I have been thinking about this subject, and I have been thinking about this subject. I have been thinking about this subject. I have been thinking about this subject. I have been thinking about this subject. I have been thinking about this subject

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

    found in oranges, which are helpful in making the kidneys clean and strong.⏎Keeps skin healthy⏎Also Check: 5 Tips on How to Find the Best Virtual Personal Trainer⏎Orange is a better source of vitamin C and consuming it can help

  2. #2 peak

    sense on how much time should be spent on a certain topic," Professor Stanley said.⏎The NSW Teachers Federation is critical of<|endoftext|>The Sun has a much higher surface temperature

  3. #3 peak

    elevation, fluid accumulation, scarring, or membrane formation.⏎Treatment for retinal disease:⏎There are a variety of<|endoftext|>In the early 90s hip-hop artist Doug E. Fresh released a single called “I-Ight (Alright)

  4. #4 peak

    up to the ice sheet surface near mountain ranges⏎How do we get to Antarctica and what is life like on the ice? Our team of six guys and two girls flew to Christchurch, New Zealand, and then were flown down to the US

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. L1:981 suppresses
  2. L1:613 suppresses
  3. L1:936 excites
  4. L0:178 suppresses
  5. L0:1224 suppresses
  6. L0:1301 weakly suppresses
  7. L1:88 weakly excites
  8. L1:173 weakly suppresses
  9. L0:2764 weakly suppresses
  10. L1:1464 weakly excites
  11. L1:1534 weakly suppresses
  12. L0:839 weakly suppresses
  13. L1:1139 weakly suppresses
  14. L0:1045 weakly excites
  15. L1:1346 weakly excites
  16. L1:451 weakly excites
  17. L1:917 weakly excites
  18. L1:719 weakly excites
  19. L1:92 weakly excites
  20. L1:853 weakly excites
  21. L0:2881 weakly suppresses
  22. L0:1644 weakly excites
  23. L1:133 weakly suppresses
  24. L1:298 weakly suppresses

Feeds later layers

  1. L3:593 excites
  2. L4:61 suppresses
  3. L6:126 suppresses
  4. L8:607 suppresses
  5. L4:41 suppresses
  6. L3:983 suppresses
  7. L8:801 suppresses
  8. L4:541 suppresses
  9. L6:265 weakly suppresses
  10. L4:241 weakly excites
  11. L6:497 weakly suppresses
  12. L14:190 weakly suppresses
  13. L12:796 weakly suppresses
  14. L5:161 weakly suppresses
  15. L3:64 weakly suppresses
  16. L10:765 weakly excites
  17. L5:107 weakly suppresses
  18. L7:975 weakly suppresses
  19. L7:220 weakly excites
  20. L7:1210 weakly suppresses
  21. L6:159 weakly suppresses
  22. L5:1 weakly suppresses
  23. L3:404 weakly excites
  24. L4:18 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 -0.79 and -0.762 values between -0.67 and -0.641 values between -0.61 and -0.581 values between -0.58 and -0.551 values between -0.55 and -0.521 values between -0.49 and -0.475 values between -0.47 and -0.449 values between -0.44 and -0.418 values between -0.41 and -0.389 values between -0.38 and -0.3518 values between -0.35 and -0.3219 values between -0.32 and -0.2940 values between -0.29 and -0.2641 values between -0.26 and -0.2378 values between -0.23 and -0.20102 values between -0.20 and -0.17147 values between -0.17 and -0.15185 values between -0.15 and -0.12247 values between -0.12 and -0.09337 values between -0.09 and -0.06482 values between -0.06 and -0.03613 values between -0.03 and 0.00629 values between 0.00 and 0.03321 values between 0.03 and 0.06213 values between 0.06 and 0.09142 values between 0.09 and 0.12105 values between 0.12 and 0.1546 values between 0.15 and 0.1847 values between 0.18 and 0.2131 values between 0.21 and 0.2315 values between 0.23 and 0.266 values between 0.26 and 0.291 values between 0.29 and 0.323 values between 0.32 and 0.352 values between 0.35 and 0.38offnegativepositiveoften

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. irma
  2. itim
  3. draws
  4. jury
  5. 发展
  6. mobil
  7. undry
  8. entions
  9. ئة
  10. ailability

Suppresses

  1. 沮
  2. Cone
  3. cone
  4. _refl
  5. .Cor
  6. abad
  7. .CompareTo
  8. 骞
  9. 編
  10. 探