Neuron Explorer

Qwen3-0.6B

'of' inside organisation and product names

A neuron in layer 7 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 word 'of' when it sits inside a formal name, like 'Office of', 'Internet of Things', 'National Institutes of Health'. It stays quiet on 'of' in ordinary phrases. Turning it up mostly made the text wander without a clear theme.

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

    in our time.⏎Melissa Rogers, a widely respected advocate for religious liberty who currently serves as special assistant to the president and executive director of the White House Office of Faith-based and Neighborhood Partnerships, also published a reflection on Religious Freedom Day.⏎

  2. #2 peak

    the only basis of implementing quality improvement. The more granular the information, the more effective it can become.⏎If the only options are in-tolerance and out-of-tolerance, quality can degrade gradually. There is no ability to determine deteriorating

  3. #3 peak

    for users to control their smart connected devices anywhere in the world, using their mobile devices. Smart connected devices can also be operated remotely using mobile applications.⏎The Internet of Things (IoT) is an upcoming technology that can be used to automate repetitive

  4. #4 peak

    However, there is still no consensus on what exactly happens when Sars-CoV-2 invades the brain. Walter J., director of the US National Institute of Neurological Disorders and Stroke in Bethesda, Maryland. Korochetz said in a

  5. #5 peak

    dramatically enhanced if these multiple electrons can be produced via a single-photon absorption event.⏎Research on carrier multiplication at Los Alamos is funded by the DOE’s Office of Basic Energy Sciences and by Los Alamos

  6. #6 peak

    , drawing thousands<|endoftext|>Healthcare is rapidly evolving into a new era, where almost everything is connected through digital technologies. This introduction of IoT, IoMT (Internet of Medical things), and medical devices have been the driving factor behind more informed care delivery

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 first one, right after the one on the net, the one at the top of the circle, and so on, or the function, $f(x) $, with a graph of a real number, and so forth. So

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

    10(MAX^2/MSE) = 10 log10(MAX/(MSE)^(1/2))^2 = 20 log10(MAX/(MSE)^(1/2)). Therefore, PSNR =

  2. #2 peak

    of the perpendiculars from X to the sides of a right angled⏎triangle. What position of X makes the length of PQ a minimum?⏎The circumcentres of four triangles are joined to form a⏎quadrilateral. What do you notice

  3. #3 peak

    scale.⏎Items you will need⏎- Calculator with statistical functions⏎Define the bel and decibel. The bel is defined mathematically as LB = log10 (P1/P0) where P1 and P0 are two quanties

  4. #4 peak

    ⏎With this bond price calculator, we aim to help you calculate the bond price issued by a government or a corporation. Finding out the current bond price is one of the most critical procedures for bond investors, as miscalculating can lead to

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. L5:1546 weakly suppresses
  2. L2:657 weakly suppresses
  3. L3:855 weakly suppresses
  4. L5:1040 weakly excites
  5. L1:2695 weakly excites
  6. L6:711 weakly suppresses
  7. L6:1465 weakly suppresses
  8. L3:1963 weakly excites
  9. L5:1154 weakly suppresses 'th' in ordinal numbers
  10. L5:424 weakly excites
  11. L5:2750 weakly suppresses
  12. L2:759 weakly suppresses
  13. L5:1620 weakly suppresses
  14. L6:357 weakly suppresses
  15. L3:2257 weakly suppresses
  16. L1:1422 weakly excites
  17. L0:630 weakly suppresses
  18. L4:2313 weakly excites
  19. L6:2027 weakly suppresses
  20. L2:589 weakly excites
  21. L6:298 weakly suppresses
  22. L5:2111 weakly suppresses
  23. L4:713 weakly suppresses
  24. L5:946 weakly suppresses

Feeds later layers

  1. L18:112 excites 'over' meaning time or duration
  2. L17:213 excites
  3. L10:1052 excites
  4. L16:177 excites zero digit inside dates and codes
  5. L20:392 excites
  6. L18:474 excites
  7. L8:318 weakly suppresses
  8. L21:381 weakly excites zeros inside long number codes
  9. L9:685 weakly suppresses
  10. L25:1654 weakly suppresses
  11. L19:2098 weakly excites
  12. L11:1712 weakly suppresses
  13. L27:882 weakly suppresses
  14. L27:2887 weakly suppresses
  15. L8:1285 weakly suppresses
  16. L22:2700 weakly suppresses
  17. L26:622 weakly excites
  18. L9:472 weakly suppresses
  19. L25:1411 weakly suppresses
  20. L20:348 weakly excites
  21. L18:1124 weakly suppresses
  22. L25:1950 weakly suppresses
  23. L19:1837 weakly excites
  24. L21:813 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 -1.21 and -1.061 values between -0.77 and -0.635 values between -0.63 and -0.4852 values between -0.48 and -0.33288 values between -0.33 and -0.191325 values between -0.19 and -0.041698 values between -0.04 and 0.10428 values between 0.10 and 0.2564 values between 0.25 and 0.3916 values between 0.39 and 0.546 values between 0.54 and 0.6910 values between 0.69 and 0.834 values between 0.83 and 0.982 values between 1.12 and 1.271 values between 1.27 and 1.411 values between 1.41 and 1.561 values between 1.85 and 2.002 values between 2.14 and 2.291 values between 2.58 and 2.721 values between 2.72 and 2.871 values between 4.47 and 4.62offnegativepositiveoften

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. adders
  2. ages
  3. adder
  4. 醯
  5. tighter
  6. UnderTest
  7. 不停的
  8. illegally
  9. tô
  10. 当他

Suppresses

  1. -default
  2. onomous
  3. minds
  4. な
  5. 幕
  6. Ầ
  7. RIX
  8. reap
  9. zer
  10. expense