Neuron Explorer

Qwen3-0.6B

"of" after nouns like level or variety

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 wakes up on the word "of" when it follows a noun such as "level", "variety" or "mode", as in "level of production". It goes quiet, or slightly negative, on "of" in phrases like "think of". Turning it up made the text wander into unrelated topics rather than anything clearly tied to this word.

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

    at all, will simply do full inclusion of a new version, or byte-wise diffing, without any attempt to determine differences at a higher level.⏎- Variety of formats. There are literally hundreds of formats for graphical images. Although there are a

  2. #2 peak

    density of the uterus, the lining of the uterus⏎Severe stomach inflammation⏎Liver and kidney toxicity⏎Changes in the level of fats in blood and signs of liver⏎There’s A LOT of evidence out there

  3. #3 peak

    used to. Due to a decrease in both the level of physical activity you partake in daily and your resting metabolic rate, loss of appetite

  4. #4 peak

    not formalized at all?⏎Frequency: how often does the innovation take place: only once, sporadically, or is it permanent or regular?⏎Mode of Selection of Participants: is the innovation open to all participants, access is restricted to

  5. #5 peak

    particular product.⏎The company wants to raise it to 40%. In order to achieve this objective the marketer has to prepare a plan in respect of the level of production and promotion efforts. It will also be decided as to who will do what

  6. #6 peak

    ups fill in the visual details (such as colors, typography, etc.) and are usually static.⏎What is fidelity in design?⏎Design fidelity refers to the level of details and functionality built into a prototype. There are three levels of fidelity: low

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 it, and I have been thinking about it again. And I have been thinking about it again again again, and I have been thinking about it again again again again again again again again again again again again again again

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

    24 hours.⏎3D printed valves⏎Once the Brescia hospital realised its regular valve supplier could not produce the valves on time, it was forced to think of alternative solutions as it had run out of breathing tubes for the intensive care machine.

  2. #2 peak

    found on beaches worldwide. When one thinks of the word recycling, it is newspapers, cardboard, glass, steel, aluminum foil

  3. #3 peak

    . If you aren’t familiar with UNIX, just think of a really complicated command language that the average person wouldn’t be able to figure out

  4. #4 peak

    the point of going as governess! What could he mean by such horrible indelicacy? To suffer her to engage herself — to suffer her even to think of such a measure!”⏎I find it interesting that at the moment of Austen

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. L26:2899 weakly suppresses
  2. L20:1648 weakly suppresses
  3. L25:1769 weakly suppresses
  4. L19:2018 weakly suppresses
  5. L26:534 weakly suppresses
  6. L22:2750 weakly suppresses
  7. L27:2778 weakly suppresses
  8. L24:2858 weakly suppresses
  9. L23:2840 weakly suppresses
  10. L23:150 weakly excites
  11. L20:1292 weakly suppresses
  12. L17:213 weakly excites
  13. L5:961 weakly suppresses 'area' and 'volume' as measures
  14. L23:2548 weakly suppresses
  15. L25:1255 weakly suppresses
  16. L3:1578 weakly suppresses
  17. L24:19 weakly excites
  18. L27:2173 weakly suppresses
  19. L27:1629 weakly suppresses
  20. L21:2804 weakly suppresses
  21. L26:2588 weakly suppresses
  22. L13:3061 weakly suppresses
  23. L27:2449 weakly suppresses
  24. L11:372 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 -1.59 and -1.522 values between -0.89 and -0.821 values between -0.82 and -0.751 values between -0.68 and -0.612 values between -0.61 and -0.545 values between -0.47 and -0.406 values between -0.40 and -0.3312 values between -0.33 and -0.2619 values between -0.26 and -0.1974 values between -0.19 and -0.12341 values between -0.12 and -0.052384 values between -0.05 and 0.02813 values between 0.02 and 0.09166 values between 0.09 and 0.1643 values between 0.16 and 0.239 values between 0.23 and 0.3010 values between 0.30 and 0.376 values between 0.37 and 0.444 values between 0.44 and 0.512 values between 0.51 and 0.582 values between 0.58 and 0.652 values between 0.65 and 0.721 values between 1.07 and 1.142 values between 1.14 and 1.21offnegativepositiveoften

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. ele
  2. ption
  3. .throw
  4. ei
  5. eo
  6. ois
  7. throw
  8. оборот
  9. ersed
  10. сла

Suppresses

  1. 这个游戏
  2. ropolitan
  3. 此基础上
  4. aker
  5. 在游戏中
  6. akers
  7. ób
  8. nown
  9. ATAR
  10. 存量