Neuron Explorer

Qwen3-0.6B

words of celebrating or promoting

A neuron in layer 17 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 leans toward words that introduce a purpose, such as 'to recognize', 'to celebrate', or 'to showcase', and 'celebration of'. It goes strongly negative on 'depend' after 'will' or 'a lot', as in 'it will depend on'. Turning it up gave rambling text with no clear link to the trigger.

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

    in Technology International Hall of Fame in a ceremony today in San Jose, California.⏎The WITI Hall of Fame was established in 1996 to recognize, honor, and promote the outstanding contributions women make to the scientific and technological communities

  2. #2 peak

    is devalued down to pennies, and often the largest use of their time - household labor - goes unpaid. IWD is a chance to remember and celebrate the struggles of women, which have won huge improvements in our

  3. #3 peak

    of data publicly available in NASA’s various mission archives. This effort is particularly appropriate for this month because April marks Global Astronomy Month, the world’s largest global celebration of astronomy.⏎The images in this quartet of galaxies represent a sample of composites

  4. #4 peak

    5.⏎SAS In-Memory Statistics Overview<|endoftext|>Thomas Edison Pitch Contest⏎The Thomas Edison Pitch Contest was created in 2010 as a competition to showcase and encourage invention, innovation and entrepreneurship among students before they get to college. Every

  5. #5 peak

    while risk is the potential for loss. Risks happen as a result of a threat exploiting a vulnerability.”⏎- Risk is the potential for loss, damage, or destruction of data caused by a cyber threat.⏎- Threat increases the likelihood of the exploitation

  6. #6 peak

    same page moneywise. This teaches your tween about the importance of communicating about money.⏎#5. Talk About Financial Options.⏎One of the hardest parts of life is encountering the financial issues that break us. Talking about the different

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 a 30000000000000000000000000000000000000000000

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 us as a population to munch our way through the equivalent of a rabbit's packed lunch on a daily basis without breaking the bank?⏎Of course, it will depend on where you shop. It's a well-known fact that large supermarket chains and

  2. #2 peak

    what happens when we are harmed, or threatened in a way which leads to trauma. It is worth saying that not all harmful events lead to trauma, a lot depends on previous experiences, support, age, neuro processing styles and a whole complex,

  3. #3 peak

    US Department of Energy on Wednesday found that solar power can quickly expand without raising electricity prices, while creating up to 1.5 million jobs.⏎But this will depend on “active cost reductions, supportive policies, and large-scale electrification,” in

  4. #4 peak

    with a high effect obstruction. It is additionally commonsense and flexible, lightweight and very savvy. It tends to be reused, however the degree of recyclability relies upon the kind of plastic. Additionally, the more food packaging is delivered, the more

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. L16:2043 suppresses
  2. L15:2508 excites
  3. L12:1229 suppresses
  4. L12:1727 excites
  5. L9:3051 suppresses
  6. L15:2989 suppresses
  7. L14:1834 excites
  8. L15:1781 excites
  9. L14:2639 suppresses
  10. L16:2945 suppresses
  11. L13:1543 excites
  12. L1:325 weakly suppresses
  13. L10:2214 weakly suppresses
  14. L15:1998 weakly excites
  15. L15:2165 weakly excites
  16. L12:1562 weakly suppresses
  17. L16:103 weakly excites
  18. L14:972 weakly suppresses
  19. L6:1784 weakly suppresses
  20. L9:1123 weakly suppresses
  21. L12:1339 weakly suppresses
  22. L11:119 weakly suppresses
  23. L8:995 weakly suppresses
  24. L6:2553 weakly excites 'for' after praise words like 'known for'

Feeds later layers

  1. L19:1808 suppresses
  2. L19:1915 excites
  3. L19:870 excites
  4. L19:2864 suppresses
  5. L18:1717 suppresses
  6. L18:1456 suppresses
  7. L20:2031 suppresses
  8. L20:941 weakly suppresses
  9. L22:2336 weakly suppresses
  10. L18:549 weakly excites
  11. L19:1361 weakly suppresses
  12. L19:2915 weakly suppresses
  13. L19:1846 weakly excites
  14. L22:1011 weakly suppresses
  15. L21:975 weakly suppresses
  16. L20:671 weakly suppresses
  17. L24:1917 weakly suppresses
  18. L23:2959 weakly suppresses
  19. L20:2325 weakly suppresses
  20. L19:2385 weakly suppresses
  21. L19:1842 weakly excites
  22. L18:1704 weakly excites
  23. L20:2530 weakly suppresses
  24. L19:1932 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 -8.69 and -8.332 values between -4.03 and -3.672 values between -3.31 and -2.954 values between -2.95 and -2.603 values between -2.60 and -2.246 values between -2.24 and -1.887 values between -1.88 and -1.5213 values between -1.52 and -1.1620 values between -1.16 and -0.80117 values between -0.80 and -0.441148 values between -0.44 and -0.092127 values between -0.09 and 0.27271 values between 0.27 and 0.63118 values between 0.63 and 0.9936 values between 0.99 and 1.3511 values between 1.35 and 1.7111 values between 1.71 and 2.062 values between 2.06 and 2.425 values between 2.42 and 2.783 values between 2.78 and 3.141 values between 5.29 and 5.65offnegativepositiveoften

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. appings
  2. ures
  3. äter
  4. ocking
  5. 悱
  6. uries
  7. ameda
  8. дов
  9. verständlich
  10. isting

Suppresses

  1. whether
  2. 人居环境
  3. Whether
  4. Ree
  5. appearance
  6. 的因素
  7. GameState
  8. Whether
  9. apus
  10. ieur