Neuron Explorer

Qwen3-0.6B

quantity word after large, high or low

A neuron in layer 0 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 fires on the word that completes a size phrase, such as 'amounts' in 'large amounts' or 'levels' in 'high levels'. Its negative side is weak and not clearly about anything. Turning it up or down did not clearly change the model's writing in a visible way.

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 the capitol city, and we are honored to be the newest member of EBR Schools.⏎IDEA believes all students can achieve at a very high-level regardless of income, geography, demographics, or special education needs. Public charter schools

  2. #2 peak

    bi-products such as saw dust and planer shavings. This production demanded large amounts of water, and thus the pump tower and a water line was established. Both

  3. #3 peak

    ons are usually produced by eating large amounts.⏎Common symptoms of jaundice include: A yellow tinge to the skin

  4. #4 peak

    , the high levels of vitamin K in broccoli can help to improve arterial function and reduce the risk of

  5. #5 peak

    low levels of H3K27me3 that could predict prognosis.⏎"We could do

  6. #6 peak

    the 1995 Galvin Report, weapons-related activities during the Cold War resulted in:⏎- About 77 million gallons of very dangerous high-level radioactive waste left over from the production of plutonium;⏎- Approximately 2.

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 number of 20000000000000000000000000000000000000000000

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 subject, there is absolutely no scientific evidence that eating shark, or taking supplements in any way, will provide any medicinal benefits whatsoever.⏎2. Swordfish photo credit: BigStock⏎This is another very large fish that eats other contaminated fish as

  2. #2 peak

    we will be doing. It's going to be another productive year in physical education class!<|endoftext|>Spring 2010 ACCESS #36⏎Just Road Pricing by Lisa Schweitzer and Brian D. Taylor⏎Economists have long advocated

  3. #3 peak

    meaning is determined" (Zuck, pp. 20). F. Biblical Theology "Biblical theology is the study of divine revelation as File Size: KB.⏎Get this from a library! Principles of biblical interpretation: (s

  4. #4 peak

    are an expectant mother, in which case, avoid these fish entirely. We will list some healthier options at the end of this list.⏎1. Shark photo credit: BigStock⏎This means any type of meat eating shark such as Longfin

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

None: this is the first layer.

Feeds later layers

  1. L6:442 excites
  2. L11:1433 excites
  3. L5:115 weakly suppresses
  4. L23:1203 weakly suppresses
  5. L1:654 weakly excites
  6. L16:1837 weakly suppresses
  7. L11:1368 weakly suppresses
  8. L14:1186 weakly excites
  9. L6:2176 weakly excites
  10. L17:1552 weakly excites
  11. L16:247 weakly suppresses
  12. L1:317 weakly suppresses
  13. L22:2178 weakly excites
  14. L1:2819 weakly excites
  15. L1:81 weakly suppresses
  16. L11:15 weakly suppresses decade and century numbers like 1950s
  17. L25:428 weakly suppresses
  18. L12:2719 weakly excites
  19. L16:2782 weakly suppresses
  20. L1:1933 weakly excites
  21. L4:1345 weakly excites
  22. L21:1909 weakly excites
  23. L2:1983 weakly excites
  24. L10:2775 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 -0.88 and -0.832 values between -0.47 and -0.424 values between -0.42 and -0.372 values between -0.37 and -0.321 values between -0.32 and -0.277 values between -0.27 and -0.2221 values between -0.22 and -0.1758 values between -0.17 and -0.12146 values between -0.12 and -0.07629 values between -0.07 and -0.022386 values between -0.02 and 0.03447 values between 0.03 and 0.09111 values between 0.09 and 0.1439 values between 0.14 and 0.1922 values between 0.19 and 0.249 values between 0.24 and 0.295 values between 0.29 and 0.346 values between 0.34 and 0.391 values between 0.39 and 0.444 values between 0.44 and 0.492 values between 0.49 and 0.541 values between 0.54 and 0.591 values between 0.64 and 0.691 values between 0.69 and 0.741 values between 0.79 and 0.841 values between 1.10 and 1.15offnegativepositiveoften

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. spb
  2. plemented
  3. ernes
  4. zej
  5. iphers
  6. .VK
  7. LLU
  8. gements
  9. osa
  10. bery

Suppresses

  1. Krank
  2. 健康成长
  3. 心理健康
  4. >*</
  5. 患病
  6. ở
  7. back
  8. .tell
  9. 健康发展
  10. 痪