Neuron Explorer

Qwen3-0.6B

word pieces inside cinnamon, China, cabar

A neuron in layer 2 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 chunk of certain words, especially cinnamon, cinnabar and place names starting with C or Ch like China and Cilicia. Turning it up made the model's writing wander without 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

    Postharvest Diseases: Phacidipycnis Rot.⏎Speck Rot in Apple⏎This rot of apples is primarily caused by

  2. #2 peak

    HIV-2.⏎The most effective extracts against HIV-1 and HIV-2, respectively, are the cinnamon bark of Cassia (Cinnamomum cassia) and the bud and fruit of the cinnamon (Card

  3. #3 peak

    slightly toxic if it comes into contact with your skin. It can cause skin allergies and forms hydrogen sulfide if combined with stomach acids.⏎Made from the mineral cinnabar, Vermilion has been used in India and China since prehistory. The

  4. #4 peak

    China. Most foods in the United States contain cassia cinnamon.⏎The distinctive smell and taste of cinnamon are due to the oily part, rich in the compound cinnamaldehyde, which is believed to be responsible for most of the health benefits

  5. #5 peak

    food bank.<|endoftext|>Some infections, however, seem to attack the mental fabric directly; intrinsic toxic processes which may be suspected on the detection of neurin and cholin in the fluids of the brain (

  6. #6 peak

    . government, has the goal of getting a vaccine approved in a far shorter time frame than a vaccine would normally take to be approved. Other countries such as Chine and Brazil are also working on creating a vaccine.<|endoftext|>2 Answers | Add Yours

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 need to be very nervous, and I can't make a decision. I need to be very nervous, and I can't make a decision. I need to be very nervous, and I can't make a decision. I need to

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

    )⏎and E is the flux in W/m²⏎More about this formula and background on the material properties in Planck, Stefan-Boltzmann, Kirchhoff and LTE.⏎The parameter called emissivity is the focus of this article.

  2. #2 peak

    this works but as the question asked for a simple answer I'll skip the math, google Maxwell's Equations and how they are used in the derivation of Kirchhoff's voltage law.⏎Batteries do make a good example for this simply

  3. #3 peak

    eat them, the cranberry shrub grows in watery bogs and has been a celebrated part of medical and culinary history. Its use among the Iroquois as a blood purifier is well documented throughout early American history though the cran

  4. #4 peak

    ! With a little bit of patience and dedication, you'll soon start to recognize the patterns and symbols and become an expert in no time.⏎A Brief On Kirchhoff S Laws With Circuit Diagram⏎Circuit Topology And Laws Applied Electricity⏎

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. L0:1301 suppresses
  2. L0:178 suppresses
  3. L0:1224 suppresses
  4. L1:1029 weakly excites
  5. L1:198 weakly suppresses
  6. L1:1099 weakly excites
  7. L1:538 weakly suppresses
  8. L1:1910 weakly excites
  9. L1:2318 weakly suppresses
  10. L1:1772 weakly suppresses
  11. L1:491 weakly excites
  12. L0:1634 weakly suppresses
  13. L0:819 weakly suppresses
  14. L0:1582 weakly suppresses
  15. L1:4 weakly suppresses
  16. L0:2413 weakly suppresses
  17. L1:2523 weakly suppresses
  18. L1:1464 weakly excites
  19. L1:2922 weakly excites
  20. L1:2648 weakly excites
  21. L1:1637 weakly excites
  22. L0:2159 weakly suppresses
  23. L0:2863 weakly suppresses
  24. L1:812 weakly suppresses

Feeds later layers

  1. L5:1889 weakly suppresses
  2. L3:2664 weakly suppresses
  3. L11:1632 weakly excites
  4. L5:355 weakly suppresses
  5. L8:306 weakly excites
  6. L15:677 weakly excites
  7. L5:1519 weakly suppresses
  8. L25:1524 weakly suppresses
  9. L22:1906 weakly suppresses
  10. L5:257 weakly excites
  11. L8:2472 weakly excites
  12. L3:2780 weakly excites
  13. L12:1025 weakly excites
  14. L25:2658 weakly suppresses
  15. L3:2839 weakly excites
  16. L8:2941 weakly excites
  17. L4:1700 weakly suppresses conservation laws and invariance
  18. L21:344 weakly suppresses
  19. L3:1978 weakly suppresses
  20. L17:2412 weakly suppresses
  21. L23:491 weakly excites
  22. L17:1941 weakly suppresses
  23. L13:437 weakly suppresses end of a short definition or claim
  24. L4:312 weakly suppresses 'that' after 'the way' or 'the first time'

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 -1.45 and -1.351 values between -1.26 and -1.162 values between -0.60 and -0.502 values between -0.50 and -0.412 values between -0.41 and -0.3121 values between -0.31 and -0.22130 values between -0.22 and -0.12910 values between -0.12 and -0.032380 values between -0.03 and 0.07357 values between 0.07 and 0.1679 values between 0.16 and 0.2616 values between 0.26 and 0.354 values between 0.35 and 0.441 values between 0.44 and 0.541 values between 1.67 and 1.771 values between 2.24 and 2.33offnegativepositiveoften

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. kl
  2. ULO
  3. funcs
  4. .Action
  5. kj
  6. �
  7. eks
  8. kh
  9. PLC
  10. shell

Suppresses

  1. дум
  2. 他们会
  3. 手腕
  4. adians
  5. arring
  6. thuật
  7. _gradients
  8. thead
  9. remembered
  10. malloc