Neuron Explorer

Qwen3-0.6B

space inside dates before the year

A neuron in layer 0 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 lights up on the blank space in dates and timestamps, just after the day number and before the year or time, as in 'Feb 23 2006'. Turning it up made the model produce number-heavy, unrelated text.

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

    intering include:⏎- Lubrication removal⏎- Oxide reduction⏎-<|endoftext|>posted on Feb, 23 2006 @ 12:29 AM⏎There are two

  2. #2 peak

    that declares a static inner class called Wheel. If we consider the real-world, we can find many objects around us,<|endoftext|>18th Jan 12 14:01<< Go back to news⏎The kinetic force of braking

  3. #3 peak

    a common enemy.⏎The effects of this split are not unlike those from the one that occurred in the<|endoftext|>Green Infrastructure for Clean Water Act⏎January 22 2010 | Know Your H20,⏎The Green Infrastructure for Clean

  4. #4 peak

    2030 as world population grows and demand for power rises, according to the 2030 water resources group:⏎Phone: +32 2 213 18 33⏎EWEA -

  5. #5 peak

    ⏎The Edcamp Model 2014 US$ 11.95 73 pages⏎The Adult Student's Guide to Survival & Success 20

  6. #6 peak

    Hildner / University of Bremen⏎The main crack<|endoftext|>analysis of algoritms⏎robert.kern at gmail.com⏎Fri Sep 10 00:10:22 CEST 2010⏎

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 how much I value the people in my life. I have been thinking about the importance of relationships and how I can be a good and kind of person in the world. I have been thinking about how much I 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

    Tucker M.R., Himmelbach A., Stein N., Mascher M., Nagel K.A., De Gara L., Simmonds<|endoftext|>Last Census - 2005⏎Cameroon is a culturally diverse costal country in Africa

  2. #2 peak

    40% lean protein, 20% fat and 40%

  3. #3 peak

    1:x in x.⏎For non-strict semantics it would be ok to store

  4. #4 peak

    3. Moving into Key Stage 2 can be a big change for both children

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. L1:1561 suppresses
  2. L9:1706 weakly suppresses
  3. L22:2631 weakly suppresses
  4. L8:749 weakly suppresses
  5. L13:966 weakly suppresses
  6. L1:2196 weakly suppresses
  7. L6:65 weakly excites
  8. L22:407 weakly excites
  9. L6:397 weakly suppresses
  10. L15:745 weakly excites
  11. L18:1369 weakly excites
  12. L18:58 weakly suppresses
  13. L13:452 weakly excites
  14. L10:95 weakly excites
  15. L11:1201 weakly excites
  16. L1:1867 weakly suppresses
  17. L22:1906 weakly suppresses
  18. L18:1761 weakly suppresses
  19. L25:1150 weakly excites
  20. L19:832 weakly excites
  21. L14:207 weakly excites
  22. L3:537 weakly excites
  23. L7:1096 weakly excites
  24. L6:736 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 -0.81 and -0.772 values between -0.54 and -0.511 values between -0.51 and -0.471 values between -0.47 and -0.433 values between -0.39 and -0.354 values between -0.35 and -0.322 values between -0.32 and -0.284 values between -0.28 and -0.245 values between -0.24 and -0.2011 values between -0.20 and -0.1645 values between -0.16 and -0.13100 values between -0.13 and -0.09273 values between -0.09 and -0.05717 values between -0.05 and -0.011696 values between -0.01 and 0.03590 values between 0.03 and 0.06235 values between 0.06 and 0.10102 values between 0.10 and 0.1467 values between 0.14 and 0.1833 values between 0.18 and 0.228 values between 0.22 and 0.251 values between 0.25 and 0.291 values between 0.29 and 0.332 values between 0.33 and 0.371 values between 0.41 and 0.441 values between 0.56 and 0.601 values between 0.63 and 0.671 values between 0.67 and 0.71offnegativepositiveoften

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. utra
  2. [tag
  3. 習
  4. читыва
  5. arine
  6. начен
  7. missions
  8. iteur
  9. .vn
  10. 过大

Suppresses

  1. 年第
  2. iltr
  3. SK
  4. tones
  5. ylation
  6. pregn
  7. Garn
  8. SY
  9. -eff
  10. 久久