Neuron Explorer

Qwen3-0.6B

name fragments split mid-word

A neuron in layer 4 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 is about parts of people's names that get split into pieces, such as the middle of a surname like "Jaip-aul" or "Todor-ov". The same kind of name piece makes it strongly negative in other cases. The pattern is clear in the examples but the direction is mixed, and steering did not show a clear theme.

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

    Research, Volume 1⏎Jonathan D. Jaipaul ✉️ Department of Environmental Studies. Faculty of Earth and Environmental Studies. Turkey

  2. #2 peak

    virus from my computer, and they refused."⏎But despite these and other experiences, Vladimir says he's determined to carry on spreading the message about HIV.⏎Jaroslav Jedlicka, AIDS Programme Manager at Prague's National Institute of Public Health,

  3. #3 peak

    select mode and select all those extra loose segmenets and delete them. Hmm it looks like a rounded pipe to me . Lets<|endoftext|>Jonathan D. Jaipaul, Devya Hemraj & Samantha Providence-Forrester⏎Published: November

  4. #4 peak

    The use and acceptance of pesticide is far from uniform. Most cities are planning a combination of removal and treatments. That includes Richfield, which is now in its fourth year of pesticide treatment, and Shoreview,

  5. #5 peak

    shocked students laughed at the blurry photograph.⏎"The jurors didn't have that reaction. They thought it was actually possible from this distance to identify a person," Todorov said. "The fact of the matter is, from this distance, you can

  6. #6 peak

    ; Nordström et al., 2017; von Arx et al., 2012). Floral humidity is created by a combination

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 question is, or none of the things?The answer is: the question is, or none of the things? Answer: the question is, or none of the things? The answer is: the question is, or none of the

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

    --- have a negative reputation that may not be deserved.⏎Nicholas DiFonzo has been studying the nature of rumors for more than 15 years. Di

  2. #2 peak

    : Pure and Applied Optics, published by the Institute of Physics in London.⏎Giulio Fanti, a professor of mechanical engineering at the University of Padua,

  3. #3 peak

    The Value of Everything: Making and Taking in the Global Economy by Mariana Mazzucato. Copyright © 2018. Available from PublicAffairs,

  4. #4 peak

    ica began under Pope Julius II, in 1506 and was completed in 1615, under Pope Paul V.⏎- Donato Bramante was to the first chief architect of St. Peter's Basilica. Many famous

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. L3:31 excites
  2. L2:1408 excites
  3. L2:2519 excites
  4. L3:658 weakly suppresses
  5. L3:2130 weakly suppresses
  6. L3:275 weakly suppresses
  7. L1:2420 weakly suppresses
  8. L2:75 weakly suppresses
  9. L3:2714 weakly excites
  10. L1:2683 weakly excites
  11. L1:1464 weakly excites
  12. L1:2160 weakly suppresses
  13. L3:2765 weakly excites
  14. L1:2244 weakly excites
  15. L2:2291 weakly excites
  16. L3:530 weakly excites
  17. L3:464 weakly excites
  18. L1:1227 weakly suppresses
  19. L1:2342 weakly excites
  20. L3:216 weakly suppresses empty slot before a figure number
  21. L1:2213 weakly suppresses
  22. L2:1803 weakly excites
  23. L0:1224 weakly suppresses
  24. L2:1 weakly suppresses

Feeds later layers

  1. L9:1086 suppresses
  2. L7:1184 suppresses
  3. L6:1067 suppresses
  4. L17:2205 suppresses
  5. L21:711 suppresses
  6. L21:751 suppresses
  7. L22:245 weakly suppresses
  8. L25:920 weakly suppresses
  9. L5:1165 weakly excites
  10. L25:97 weakly suppresses
  11. L27:1352 weakly excites
  12. L23:352 weakly suppresses
  13. L26:637 weakly suppresses
  14. L26:869 weakly excites word before him or he
  15. L27:2018 weakly suppresses
  16. L27:929 weakly suppresses
  17. L9:2098 weakly suppresses
  18. L5:2384 weakly suppresses
  19. L25:555 weakly excites
  20. L22:654 weakly suppresses
  21. L10:1826 weakly suppresses
  22. L11:416 weakly suppresses
  23. L7:598 weakly suppresses
  24. L20:1251 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 -1.60 and -1.511 values between -1.43 and -1.353 values between -1.10 and -1.011 values between -0.68 and -0.602 values between -0.60 and -0.515 values between -0.51 and -0.4313 values between -0.43 and -0.3537 values between -0.35 and -0.26143 values between -0.26 and -0.18410 values between -0.18 and -0.10937 values between -0.10 and -0.011126 values between -0.01 and 0.07744 values between 0.07 and 0.16320 values between 0.16 and 0.24118 values between 0.24 and 0.3234 values between 0.32 and 0.416 values between 0.41 and 0.492 values between 0.49 and 0.571 values between 0.66 and 0.741 values between 0.74 and 0.821 values between 1.16 and 1.241 values between 1.24 and 1.321 values between 1.66 and 1.74offnegativepositiveoften

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. ortion
  2. orgh
  3. uggested
  4. buried
  5. ác
  6. ROWN
  7. 用地
  8. {}).
  9. speech
  10. ency

Suppresses

  1. |--
  2. 微信号
  3. zza
  4. 经查
  5. Rosenstein
  6. isos
  7. 彻
  8. estring
  9. 已经超过
  10. смотреть