Neuron Explorer

Qwen3-0.6B

content nouns in health and advice text

A neuron in layer 13 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 key noun in a phrase in medical or how-to writing, such as 'references', 'treatment options' or 'numbers'. Its strongest negative cases are similar nouns in advice text, like 'changes' or 'tasks'. Turning it up did not produce a clear theme in the model's writing.

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

    Approximate location of Wernicke's area highlighted in gray|⏎| This article needs additional citations for verification.|⏎Please help improve this article by adding reliable references. Unsourced material may be challenged and removed. (October 200

  2. #2 peak

    . Clearly there was no consensus on the definition of a microservice and what it actually means. It is quite a buzzword these days but none of the panelists could come to a common understanding. This raised a few questions in my head.

  3. #3 peak

    irritation and damage to the ocular surface. There are many different forms of dry eye, each with a unique cause, and each with a unique set of treatment options available. The following video provides a brief introduction to dry eye disease, symptoms of

  4. #4 peak

    determine the cause and get you appropriate treatment.⏎Here are some of the medical conditions that can cause dizziness and lightheadedness, and how they’re treated:⏎Benign Paroxysmal Positional Vertigo (BPPV)⏎

  5. #5 peak

    modifications to the original file in the meantime. So your modified file could not replace it.⏎That is why patches exist. Patches list the modifications, the line numbers and a few other useful information to be able to put that patch back into the

  6. #6 peak

    few seconds:⏎See how “normal” is translated from English to French with more examples in context⏎- Create your own vocabulary list⏎- Contribute to the Collaborative Dictionary

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 just need to make sure that the system is designed to change the behavior of the customers that are in the present day that would be based on the future, and I am assuming the current system will be in the future that will affect 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

    to be.⏎With just a few simple tweaks to your ways of living, you can easily kick-start your sustainable lifestyle. It’s best to start with small changes so that they stick and build it up from there.⏎So how can you start

  2. #2 peak

    ing the issue quickly, ensuring timely and accurate repairs or replacements.⏎The Importance of Regular Maintenance⏎Maintenance is vital for the longevity of HVAC systems. This includes routine tasks such as replacing air filters, cleaning coils, inspecting ductwork, and ensuring

  3. #3 peak

    at which electrons flow past a point in a complete electrical circuit.⏎Want to Slash Your EMF Health Risks?⏎Good! Learn the one small change you should make right now.⏎To understand this better, think of electric current as athletes on a

  4. #4 peak

    Brooklyn, Iowa. A lot of kids work for neighbors, and there are certain things parents would like their kids to be able to do when they reach age 14 or 15—including running a tractor and

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. L10:2456 suppresses
  2. L7:1815 suppresses
  3. L12:2731 suppresses
  4. L12:1849 suppresses
  5. L10:910 excites
  6. L12:1873 excites
  7. L10:3016 suppresses
  8. L10:882 suppresses
  9. L11:2298 suppresses
  10. L11:1024 suppresses
  11. L9:1066 excites
  12. L12:2247 weakly suppresses
  13. L3:1407 weakly suppresses
  14. L11:2765 weakly excites
  15. L4:1052 weakly excites
  16. L8:69 weakly excites
  17. L12:1427 weakly suppresses
  18. L12:2604 weakly excites
  19. L12:2246 weakly suppresses
  20. L1:2942 weakly suppresses
  21. L7:910 weakly excites
  22. L3:3060 weakly suppresses
  23. L12:1475 weakly suppresses
  24. L4:2095 weakly suppresses

Feeds later layers

  1. L16:2161 suppresses
  2. L14:2069 excites
  3. L16:2700 suppresses
  4. L19:2122 suppresses
  5. L14:492 suppresses
  6. L21:1579 suppresses
  7. L15:1691 suppresses
  8. L15:1851 suppresses 'that' after a noun, 'being' in passives
  9. L17:2504 weakly suppresses
  10. L18:2141 weakly suppresses
  11. L19:2584 weakly suppresses
  12. L14:2462 weakly excites
  13. L17:1222 weakly suppresses
  14. L19:1233 weakly excites
  15. L15:741 weakly excites
  16. L14:2835 weakly suppresses
  17. L20:1441 weakly suppresses
  18. L16:2474 weakly suppresses
  19. L18:1691 weakly suppresses
  20. L19:226 weakly excites
  21. L16:1858 weakly suppresses
  22. L14:2791 weakly suppresses
  23. L16:531 weakly suppresses
  24. L18:1022 weakly excites

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 -4.21 and -4.021 values between -2.85 and -2.661 values between -2.46 and -2.272 values between -2.27 and -2.071 values between -2.07 and -1.882 values between -1.69 and -1.496 values between -1.49 and -1.3019 values between -1.30 and -1.1015 values between -1.10 and -0.9122 values between -0.91 and -0.7148 values between -0.71 and -0.5288 values between -0.52 and -0.32369 values between -0.32 and -0.132401 values between -0.13 and 0.06698 values between 0.06 and 0.26140 values between 0.26 and 0.4561 values between 0.45 and 0.6512 values between 0.65 and 0.8412 values between 0.84 and 1.046 values between 1.04 and 1.232 values between 1.43 and 1.621 values between 3.37 and 3.56offnegativepositiveoften

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. rido
  2. adero
  3. ercul
  4. equality
  5. CCI
  6. Layout
  7. theless
  8. 实
  9. Doch
  10. 谁

Suppresses

  1. 比如
  2. inine
  3. ork
  4. 比如说
  5. 譬如
  6. 例如
  7. AppState
  8. daq
  9. 诸如
  10. -serving