Neuron Explorer

Qwen3-0.6B

unclear: word stems in technical text

A neuron in layer 15 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 fires on the middle pieces of longer technical or specialist words, such as 'roasting', 'math', 'orthodontist' and 'writing', in informational articles. It goes negative on words like 'only' and 'this' after 'to do' or 'can'. Turning it up did not produce a recognisable theme in the model's replies.

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

    , with thousands of compounds being created and broken down. Green coffee is loaded into the heated roasting chamber and undergoes drying as the first stage of the roasting cycle. During the first few minutes, the beans desiccate most intensively

  2. #2 peak

    crawl, stand and eventually walk.⏎Why is tummy time important?⏎Infants spend a lot of time lying on their back, especially a few weeks after being born. If they’re not lying on their backs, they are being carried or sw

  3. #3 peak

    to Graphs⏎A formal introduction to graphs may be found in most discrete math textbooks. The purpose of this section is not to provide such an introduction. Rather

  4. #4 peak

    /dev and /devices directories in response to dynamic reconfiguration events or file system accesses are handled by devfsadmd, the daemon version of the devfsadm command. This daemon is started by the service management facility when a system is bo

  5. #5 peak

    on a TV monitor) to perform the surgical procedure. These laparoscopic procedures offer benefits to the conventional open surgery including:⏎- Reduced Risk of Infection⏎Single Incision laparoscopic surgery

  6. #6 peak

    Care of Your Teeth with Braces⏎When you have braces, you'll have to be very good about following the brushing and flossing instructions your orthodontist gives you.⏎You can't eat hard, sticky or gooey foods such as

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: to have a set of information, I need to know the information, I need to know that, I need to know that, I need to know that, I need to know that, I need to know that... (the information,

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

    ism. We must remind them of what they knew as toddlers: mistakes are a valuable part of the learning process. Innovation and discovery can only be fostered if we give our children the freedom to fail. Schools can build

  2. #2 peak

    in the right hands. It can determine the age, position and heart rate of the foetus, along with an indication of its overall health. But to do this consistently can take many years of practice.⏎Meanwhile, in developing countries, “a

  3. #3 peak

    kinds of deception are supported by visual information?⏎These are the types of questions we are attempting to answer as we build our typology of visual misinformation. To do so, we are using the data from the CrossCheck French election project and First Draft

  4. #4 peak

    <= R5+1; R6 <= R6-R3 would be held up in a non-scheduled pipeline, because the second operation based on R2 has to wait for the load from memory into R1, thus halting the CPU

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. L7:2637 suppresses
  2. L12:1522 excites comma after a 'before...' clause
  3. L8:2965 suppresses
  4. L13:2948 suppresses
  5. L11:2506 excites
  6. L12:3018 excites
  7. L13:487 weakly excites
  8. L14:835 weakly suppresses
  9. L11:1234 weakly excites
  10. L11:2107 weakly suppresses
  11. L12:1496 weakly suppresses
  12. L13:1256 weakly suppresses
  13. L11:1986 weakly suppresses
  14. L14:3019 weakly excites
  15. L14:2633 weakly suppresses
  16. L12:2909 weakly suppresses
  17. L12:268 weakly excites
  18. L4:195 weakly excites
  19. L11:1825 weakly excites
  20. L8:1006 weakly suppresses
  21. L14:174 weakly excites
  22. L11:2004 weakly suppresses
  23. L10:2681 weakly excites
  24. L13:533 weakly excites

Feeds later layers

  1. L20:2888 suppresses
  2. L20:2529 suppresses
  3. L18:2494 suppresses
  4. L18:96 suppresses
  5. L19:2249 suppresses
  6. L17:1347 suppresses
  7. L17:2521 suppresses
  8. L17:863 suppresses
  9. L18:1563 suppresses
  10. L18:2441 suppresses
  11. L16:3061 suppresses
  12. L16:2788 suppresses
  13. L17:381 suppresses
  14. L18:1932 suppresses
  15. L16:2284 suppresses
  16. L18:2514 suppresses
  17. L20:2134 suppresses
  18. L24:1041 suppresses
  19. L17:1113 suppresses
  20. L22:1336 suppresses
  21. L16:514 weakly suppresses
  22. L18:1003 weakly suppresses
  23. L19:2139 weakly suppresses
  24. L21:2805 weakly suppresses 'the' right before 'subject' or 'source'

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 -3.66 and -3.531 values between -3.27 and -3.141 values between -2.62 and -2.492 values between -1.57 and -1.441 values between -1.18 and -1.051 values between -1.05 and -0.926 values between -0.92 and -0.795 values between -0.79 and -0.668 values between -0.66 and -0.5341 values between -0.53 and -0.40123 values between -0.40 and -0.27367 values between -0.27 and -0.141120 values between -0.14 and -0.011598 values between -0.01 and 0.12447 values between 0.12 and 0.25145 values between 0.25 and 0.3828 values between 0.38 and 0.515 values between 0.51 and 0.653 values between 0.65 and 0.782 values between 0.78 and 0.912 values between 0.91 and 1.041 values between 1.43 and 1.56offnegativepositiveoften

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. aurant
  2. otify
  3. anguard
  4. inally
  5. ummings
  6. ngạc
  7. etCode
  8. ington
  9. .getResult
  10. .HttpServletResponse

Suppresses

  1. 的前提
  2. requires
  3. require
  4. requires
  5. require
  6. Requires
  7. 都需要
  8. Require
  9. 需要
  10. prerequisites