Neuron Explorer

Qwen3-0.6B

'to' starting a purpose clause

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 fires on the word 'to' when it introduces a reason or goal, as in 'click the link to confirm' or 'add water to maintain'. It fires less on other uses of 'to'. Turning it up or down did not make the writing about anything in particular.

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

    related to this resource.⏎You have successfully created an account.⏎A verifications link was sent to your email at . Please check your email and click on the link to confirm your email address and fully activate your iCPALMS account. Please check

  2. #2 peak

    2 quarts 3% hydrogen peroxide to a tub of warm water. Soak at least 1/2 hour, adding hot water as needed to maintain a comfortable water temperature.⏎To cure a foot fungus, simply spray a

  3. #3 peak

    sowing, to redirect the plants' energy to the roots. This will ensure long-term health and better flowering.⏎- After this, cover them with mulch to prevent weeds and preserve moisture.⏎- Make sure you fertilize marig

  4. #4 peak

    ensuring they match the model and capacity of the system.⏎Energy Efficiency Considerations⏎Energy efficiency is a critical consideration in HVAC systems. Opting for energy-efficient parts can significantly reduce the environmental impact and lower operating costs. When selecting parts, look for

  5. #5 peak

    pile is getting too dry, you can add water to it to help keep everything moist and speed up decomposition.⏎One way to speed up the compost

  6. #6 peak

    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 you continue to use this website without

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 can't, I can't, I can't, I can't, I can't, I can't, I can't, I can't, I can't, I can't, I can't, I can't,

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

    13.5 out of 15 as a percent, you’d enter a 13.5 out of 15 into a calculator. The calculator will do the rest, including converting the decimal to a percentage.⏎To find the

  2. #2 peak

    to Muhammad, such as:⏎“There is no disease that Allah has created, except that He also has created its treatment.⏎“Make use of

  3. #3 peak

    be tweaked with a few mouseclicks. The 3D printer can run unattended, and can make many things which are too complex for a traditional factory to handle. In time, these amazing machines may be able to make almost anything,

  4. #4 peak

    is often diagnosed during a visit to your doctor. Blood pressure is measured using a cuff around your arm and a device called a sphygmomanometer. Your doctor may ask you to sit quietly for five minutes before checking your blood pressure.⏎If

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. L13:685 strongly suppresses
  2. L13:2681 excites
  3. L6:2000 suppresses
  4. L6:904 suppresses
  5. L12:2212 suppresses 'to' or 'for' opening a purpose phrase
  6. L11:896 excites comma closing an introductory phrase
  7. L8:94 excites
  8. L13:1237 excites
  9. L13:504 excites
  10. L9:1890 excites
  11. L6:439 suppresses
  12. L7:681 excites 'from' after prevent or stop
  13. L6:2747 weakly suppresses
  14. L12:1684 weakly suppresses comma after a transition word
  15. L6:1880 weakly excites "by" introducing a deadline or means
  16. L12:3011 weakly suppresses
  17. L9:661 weakly excites
  18. L13:527 weakly suppresses
  19. L6:2372 weakly excites
  20. L13:1320 weakly excites
  21. L1:906 weakly suppresses
  22. L12:2704 weakly suppresses
  23. L12:2279 weakly suppresses
  24. L12:804 weakly excites

Feeds later layers

  1. L15:1543 excites
  2. L15:1176 excites
  3. L16:559 excites weak, mixed: verbs and adverbs in claims
  4. L17:472 excites
  5. L24:208 excites
  6. L20:241 excites
  7. L15:233 excites
  8. L19:2040 excites
  9. L19:617 excites
  10. L20:922 excites
  11. L18:391 excites
  12. L19:544 suppresses
  13. L19:2432 excites
  14. L23:1169 excites
  15. L24:442 excites
  16. L22:1166 excites 'to' after words needing an infinitive
  17. L17:2008 excites
  18. L17:616 excites verb right after a long subject
  19. L19:271 excites
  20. L16:2762 excites
  21. L22:459 excites
  22. L21:123 excites
  23. L18:472 excites
  24. L17:530 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 -2.79 and -2.592 values between -1.80 and -1.605 values between -1.01 and -0.8143 values between -0.81 and -0.61203 values between -0.61 and -0.42687 values between -0.42 and -0.221369 values between -0.22 and -0.021219 values between -0.02 and 0.18255 values between 0.18 and 0.3859 values between 0.38 and 0.5717 values between 0.57 and 0.7712 values between 0.77 and 0.977 values between 0.97 and 1.174 values between 1.17 and 1.376 values between 1.37 and 1.563 values between 1.56 and 1.763 values between 1.76 and 1.961 values between 2.16 and 2.352 values between 2.35 and 2.551 values between 2.55 and 2.751 values between 2.75 and 2.953 values between 2.95 and 3.152 values between 3.34 and 3.541 values between 3.54 and 3.741 values between 3.94 and 4.141 values between 4.93 and 5.13offnegativepositiveoften

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. ei
  2. ży
  3. ube
  4. create
  5. /forum
  6. YTE
  7. 素
  8. 使我
  9. يج
  10. IRD

Suppresses

  1. ingly
  2. 起来了
  3. движ
  4. 奥林
  5. Enough
  6. issued
  7. 想到了
  8. ']):
  9. Fortunately
  10. 少不了