Neuron Explorer

Qwen3-0.6B

'to' starting a purpose clause

A neuron in layer 21 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 after an action, as in 'check the website to request a copy' or 'download the rubric to understand'. Turned up, the model's output drifted toward the word 'understanding', which fits this purpose sense.

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

    . Government employees on official time, and is therefore in the public domain.⏎- This publication may be available in hard copy. Check the Northern Research Station web site to request a printed copy of this publication.⏎- Our on-line publications are scanned and

  2. #2 peak

    questions give students a sneak preview of what to expect on the test. Furthermore, students can look at the results of a practice test to determine which skills they need to work on.⏎Some students may do well on the

  3. #3 peak

    schools.⏎- Check the First Amendment Websites to ensure that all are appropriate for us in your class. If desired, make an

  4. #4 peak

    your doctor or pharmacist if you have any questions. You can also visit the Food and Drug Administration (FDA) website ( Web Site) or the manufacturer's website to obtain the Medication<|endoftext|>Gingivitis, periodontitis, and gum

  5. #5 peak

    5: Assignment 1 Dropbox.⏎Grading Criteria and Rubric⏎Note: This assignment will be graded using a rubric. Download and read the rubric to understand the expectations.<|endoftext|>A new bionic hand has been developed that may give

  6. #6 peak

    until two years after they were spawned when they are large enough to be harvested and sold to restaurants around New York.⏎Check out Blue Ocean’s ranking for Oysters to learn more.<|endoftext|>Urban and Rural Synergies⏎Jazz benefited greatly from

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: when you're in the middle of a group, you're not giving yourself time to give one full and complete one-on-one conversation, but rather you're giving yourself time to provide a group by giving one group by one. So, that

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

    These pocket-sized books are the perfect way to get ahead in a new subject quickly. Our expert authors combine facts, analysis, perspective, new ideas, and enthusiasm to make interesting and challenging topics highly readable.<|endoftext|>The MPI Multidimensional Poverty Index

  2. #2 peak

    We hold these Truths to be self-evident, that all Men are created equal, that they are

  3. #3 peak

    ile not only demonstrates Bradbury's incredibly vivid style, but also connects the house to our own flesh and blood. By describing the house in human terms, the author hopes the reader will identify with it, and thus feel empathy for the idea that

  4. #4 peak

    .⏎4.-Do you participate in any educational or social activity that promotes STEM careers in girls?⏎Throughout my career I have been involved in outreach events aimed at school children. However never specifically aimed at girls, I believe it is important 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. L20:810 suppresses
  2. L19:1769 excites
  3. L18:2760 excites 'to' after a verb of sharing or using
  4. L20:1124 suppresses
  5. L19:2632 suppresses
  6. L19:2535 excites
  7. L20:592 suppresses
  8. L19:2264 suppresses
  9. L20:1015 excites
  10. L19:1418 weakly excites
  11. L20:2369 weakly excites
  12. L19:3052 weakly suppresses
  13. L16:1153 weakly suppresses
  14. L17:2387 weakly excites
  15. L0:1280 weakly excites
  16. L19:386 weakly suppresses
  17. L20:585 weakly excites
  18. L17:2856 weakly suppresses
  19. L19:3044 weakly suppresses
  20. L19:2460 weakly excites
  21. L19:3048 weakly excites
  22. L11:2420 weakly suppresses
  23. L20:813 weakly suppresses
  24. L12:2212 weakly suppresses 'to' or 'for' opening a purpose phrase

Feeds later layers

  1. L22:3009 excites
  2. L24:1996 suppresses
  3. L25:1065 excites
  4. L25:640 suppresses
  5. L22:952 suppresses
  6. L24:1202 excites
  7. L22:1338 suppresses
  8. L25:827 excites
  9. L22:1896 excites
  10. L24:963 suppresses
  11. L24:510 suppresses
  12. L23:2824 suppresses
  13. L27:1408 excites
  14. L27:1274 suppresses
  15. L23:2098 weakly excites
  16. L22:1478 weakly suppresses
  17. L22:2002 weakly excites
  18. L23:2409 weakly suppresses
  19. L23:1827 weakly excites
  20. L25:958 weakly suppresses
  21. L24:769 weakly excites
  22. L27:1284 weakly excites
  23. L24:1229 weakly excites
  24. L24:1302 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 -12.99 and -12.431 values between -11.86 and -11.291 values between -10.73 and -10.161 values between -10.16 and -9.602 values between -9.60 and -9.031 values between -7.90 and -7.332 values between -7.33 and -6.773 values between -6.77 and -6.204 values between -6.20 and -5.637 values between -5.63 and -5.071 values between -5.07 and -4.505 values between -4.50 and -3.9410 values between -3.94 and -3.3714 values between -3.37 and -2.8019 values between -2.80 and -2.2432 values between -2.24 and -1.6750 values between -1.67 and -1.11161 values between -1.11 and -0.541686 values between -0.54 and 0.031755 values between 0.03 and 0.59109 values between 0.59 and 1.1621 values between 1.16 and 1.729 values between 1.72 and 2.293 values between 2.29 and 2.862 values between 2.86 and 3.422 values between 3.42 and 3.992 values between 3.99 and 4.552 values between 6.25 and 6.821 values between 6.82 and 7.381 values between 9.08 and 9.65offnegativepositiveoften

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. 了解
  2. 瞭解
  3. understand
  4. 了解到
  5. 理解
  6. understanding
  7. 得知
  8. understands
  9. 知道了
  10. 找到

Suppresses

  1. 提供
  2. give
  3. giving
  4. provide
  5. 告诉我们
  6. gives
  7. 给我们
  8. providing
  9. 提供了
  10. 告诉他