Neuron Explorer

Qwen3-0.6B

'is' after a topic noun phrase

A neuron in layer 14 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 'is' (or 'are') right after a phrase that names a topic, like 'One of the most common uses of this calculation is' or 'The exception to this is'. It is waiting for the sentence's main point to come next. Turning it up made the model's writing wander into generic, repetitive phrases about 'the most important thing'.

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

    , but the⏎site also needs to let engineers access the information they need⏎in as few clicks as possible. The one component engineers have in⏎short supply is time,” shares Kevin Hess, Mouser

  2. #2 peak

    copyright of the author of that content. To comply with copyright law, you must receive permission from the copyright holder before you download any content. The exception to this is Fair Use. As in copying printed material such as books in the library, you

  3. #3 peak

    % is the same as 25/100 and equivalent to 0.25 . One of the most common uses of this calculation is when purchasing a product on sale.⏎Thus, if the marked price is 1

  4. #4 peak

    yoga) may also be considered. The other common treatment option is surgery, which consists of two different types: open release and endoscopic surgery.⏎

  5. #5 peak

    of the main concerns about PET bottles is the potential for chemicals to leach from the plastic into the contents of the bottle

  6. #6 peak

    switch that can be controlled at room temperature and that will hold its position without the need for any electricity at all. The only real downside to it thus far, is that it takes far more energy to make the switch move than does conventional non-m

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: and the most important thing to the most important thing to the most important thing to the most important thing to the most important thing to the most important thing to the most important thing to the most important thing to the most important thing to the most

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

    army units as a pastime when not at war – the process of capturing the beasts is recorded in several Roman mosaics.⏎The cookie settings on this website are set to "allow cookies" to give you the best browsing experience possible. If

  2. #2 peak

    peaceful world. But if the war continues, the world will be destroyed. War occurs when countries are not satisfied with what they have. They always want more which is a necessary to them. Their wish to rule other countries and widen their empires

  3. #3 peak

    .⏎What Is Thrombosis?⏎Thrombosis is when the body forms a blood clot inside a blood vessel. A blood clot in a blood vessel is also called a thrombus. A thrombus can stop blood flow to parts of

  4. #4 peak

    and they are the lock and key model and induced fit hypothesis.⏎- The shape of the active site is determined primarily by the tertiary structure of the protein, which is why the bonding and interactions between the R groups of the amino acids must be correct

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. L13:897 excites
  2. L6:2237 excites
  3. L6:1567 excites colon introducing a summary point
  4. L9:97 excites
  5. L13:948 suppresses
  6. L6:2572 weakly excites
  7. L12:1222 weakly suppresses
  8. L3:367 weakly excites
  9. L13:1129 weakly excites
  10. L2:884 weakly excites
  11. L10:272 weakly suppresses
  12. L13:287 weakly suppresses
  13. L11:1667 weakly excites
  14. L8:376 weakly excites
  15. L1:444 weakly suppresses
  16. L11:2733 weakly excites
  17. L13:612 weakly suppresses
  18. L9:84 weakly suppresses
  19. L12:850 weakly suppresses
  20. L3:56 weakly excites
  21. L1:644 weakly suppresses
  22. L3:1043 weakly suppresses
  23. L9:1059 weakly excites
  24. L9:492 weakly suppresses

Feeds later layers

  1. L17:1171 suppresses
  2. L16:2790 excites
  3. L17:2380 excites
  4. L16:1535 excites
  5. L16:2743 suppresses
  6. L22:3024 weakly suppresses
  7. L17:143 weakly excites
  8. L18:194 weakly excites
  9. L20:2026 weakly suppresses
  10. L16:1739 weakly suppresses
  11. L21:2473 weakly suppresses
  12. L17:2684 weakly excites
  13. L15:1299 weakly excites
  14. L16:1267 weakly excites
  15. L19:2730 weakly suppresses
  16. L24:1650 weakly excites
  17. L16:316 weakly excites
  18. L16:2067 weakly excites
  19. L16:2444 weakly suppresses
  20. L17:156 weakly suppresses
  21. L17:218 weakly excites
  22. L15:1851 weakly suppresses 'that' after a noun, 'being' in passives
  23. L18:92 weakly excites
  24. L17:255 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.

5 values between -1.28 and -1.111 values between -1.11 and -0.934 values between -0.93 and -0.7515 values between -0.75 and -0.5888 values between -0.58 and -0.40422 values between -0.40 and -0.221229 values between -0.22 and -0.051608 values between -0.05 and 0.13396 values between 0.13 and 0.3187 values between 0.31 and 0.4814 values between 0.48 and 0.669 values between 0.66 and 0.844 values between 0.84 and 1.026 values between 1.02 and 1.192 values between 1.19 and 1.377 values between 1.37 and 1.551 values between 1.72 and 1.901 values between 1.90 and 2.082 values between 2.08 and 2.251 values between 2.25 and 2.431 values between 2.61 and 2.781 values between 2.78 and 2.961 values between 3.31 and 3.492 values between 4.02 and 4.201 values between 5.61 and 5.79offnegativepositiveoften

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. 那就是
  2. ancias
  3. 的就是
  4. 一个是
  5. (express
  6. Orden
  7. :start
  8. →Schema
  9. Orden
  10. ElementType

Suppresses

  1. ipur
  2. �
  3. hevik
  4. 最长
  5. Islamabad
  6. ormal
  7. LGBT
  8. ấu
  9. تد
  10. nier