Neuron Explorer

Qwen3-0.6B

the word 'social'

A neuron in layer 18 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 responds to the word 'social', most often in phrases like 'social media', 'social indicators' or 'social exchange'. It is almost silent otherwise. Turning it up made the text drift toward social and economic topics only occasionally.

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

    different parts of the country facing similar challenges⏎- Stay ‘tuned-in’ to new priorities emerging that might affect your project⏎- Stay engaged with the wider social marketing community and the growing stock of

  2. #2 peak

    critical to closing the digital divide, and essential for boosting economic growth. Internet access allows women to gain a voice, seek information, or expand their networks beyond existing social boundaries.⏎Yet, African governments are failing to spend funds collected specifically to connect the

  3. #3 peak

    lot of time on their devices. With more free time at your disposal, evenings are a seemingly perfect opportunity to scroll through your phone and catch up on the latest social media posts. But do you know that using devices before bed can have a detrimental

  4. #4 peak

    challenges in space security.⏎By entering this website, you consent to the use of technologies, such as cookies and analytics, to customise content, advertising and provide social media features. This will be used to analyse traffic to the website, allowing us

  5. #5 peak

    change will disproportionately affect California’s most disadvantaged and isolated communities, according to a recent report from the Pacific Institute.⏎By looking at a broad array of factors – from social indicators such as income and birth rates, to environmental ones such as tree cover and

  6. #6 peak

    cess, but they also function as a platform for social exchange. On weekends, funeral attire often dominates the public space. The clothing styles

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: what is the best way to get a good first line in a game is not to have an attack, but to have what they have to have to get a good first line is the first what they have to have what they have to have

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

    This preview has intentionally blurred sections. Sign up to view the full version.View Full Document⏎Unformatted text preview: itself, and if ( ) ( ) lim lim x c x c h

  2. #2 peak

    –2. Sign up to view the full content.⏎This preview has intentionally blurred sections. Sign up to view the full version.View Full Document⏎Unformatted text preview: 60 | | | | CHAPTER 1 FUNCTIONS AND MODELS Not

  3. #3 peak

    of the moon. Its mission? To take pictures of this unexplored part of the moon, collect soil samples, and plant a mini garden. Awesome, right? So how successful has the Chang’e

  4. #4 peak

    place.⏎Ready to start? Let’s dive in!⏎A little history of the SSL certificate⏎They say “to know your future you must know your past”, right? Put into a closer, business perspective, I’d rephrase that with “

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. L16:241 weakly suppresses
  2. L17:337 weakly excites
  3. L6:16 weakly excites
  4. L16:332 weakly excites plural '(s)' in parentheses
  5. L15:785 weakly suppresses 'of' in set phrases like 'of course'
  6. L16:119 weakly excites
  7. L17:776 weakly excites
  8. L16:126 weakly suppresses
  9. L11:538 weakly excites
  10. L16:305 weakly suppresses
  11. L15:39 weakly suppresses
  12. L7:1614 weakly suppresses
  13. L16:359 weakly excites
  14. L7:122 weakly excites
  15. L2:3010 weakly excites
  16. L5:46 weakly excites
  17. L17:2812 weakly excites
  18. L0:2026 weakly suppresses
  19. L2:1755 weakly suppresses
  20. L3:1356 weakly excites
  21. L10:1305 weakly excites
  22. L17:542 weakly excites
  23. L16:142 weakly excites
  24. L0:2493 weakly suppresses

Feeds later layers

  1. L19:889 weakly excites
  2. L20:1684 weakly suppresses
  3. L21:2595 weakly suppresses
  4. L26:1592 weakly excites
  5. L21:2778 weakly excites
  6. L19:574 weakly excites
  7. L26:2441 weakly suppresses
  8. L26:1983 weakly suppresses the 'lo' fragment in loom and loaves
  9. L23:1135 weakly excites
  10. L24:2036 weakly suppresses
  11. L24:677 weakly suppresses
  12. L19:2189 weakly suppresses
  13. L20:792 weakly excites
  14. L19:2317 weakly suppresses
  15. L20:2383 weakly excites
  16. L24:508 weakly excites
  17. L26:1517 weakly excites
  18. L24:281 weakly suppresses
  19. L19:1001 weakly excites "be" after will or you'll
  20. L25:488 weakly excites
  21. L25:2726 weakly suppresses
  22. L26:2044 weakly suppresses
  23. L27:493 weakly suppresses
  24. L21:2523 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 -4.05 and -3.171 values between -3.17 and -2.292 values between -2.29 and -1.4083 values between -1.40 and -0.523666 values between -0.52 and 0.36141 values between 0.36 and 1.245 values between 1.24 and 2.132 values between 2.13 and 3.011 values between 3.89 and 4.771 values between 4.77 and 5.661 values between 6.54 and 7.421 values between 7.42 and 8.301 values between 21.54 and 22.431 values between 28.60 and 29.491 values between 30.37 and 31.25offnegativepositiveoften

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. vement
  2. 蠊
  3. uated
  4. GN
  5. __',
  6. entar
  7. |-
  8. aces
  9. tempt
  10. venes

Suppresses

  1. 对自己的
  2. ight
  3. 对未来
  4. خت
  5. http
  6. 都会有
  7. 什么是
  8. lee
  9. 什么叫
  10. ération