Neuron Explorer

Qwen3-0.6B

"here" in web link phrases

A neuron in layer 8 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 "here" when it ends a call to action like "click here" or "available here", the kind of link text found on web pages. Turning it up gave output that did not clearly follow this pattern.

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

    being funded by its wealth.⏎Scottish Webcams HERE North American Webcams HERE World Webcams HERE<|endoftext|>E. Cobham Brewer 18101

  2. #2 peak

    Ltd.<|endoftext|>British Poetry And Drama Unit 2 Jacobean Drama Notes, College and University Answer Bank of Education for BA, and Post Graduate Notes and Guide Available here, Solutions of British Poetry And Drama Unit 2 Jacobean Drama to each Unit

  3. #3 peak

    treatments listed in sources for treatment of Apraxia may include:⏎Review further information on Apraxia Treatments.⏎Real-life user stories relating to Apraxia:⏎Some of the comorbid or associated medical symptoms for Apraxia may include

  4. #4 peak

    of how such injuries occur.⏎Check out our article on the phases of whiplash here.⏎To read more about the typical symptoms of whiplash, click here⏎For more info about recovery and treatment strategy and options, click here⏎6

  5. #5 peak

    create a new variable and:⏎Now with the List data type, things haven’t changed much. To create a List of Users, you create a new variable and:⏎The awesome part is that now to access the user’s name

  6. #6 peak

    , however, gained only 0.3 years over the same period, to 75.4. (For a table listing all the states, click here.)⏎Lichtenberg measured a number of variants that might explain this gap, including

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: does a person who doesn't own a pet dog not own a cat or a mouse? ...? The question is whether the removal of the specific element from the overall composition is allowed to be retained? ...? The question is whether the removal

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

    , a behavior that is considered ethical or socially acceptable in one community may be utterly unacceptable<|endoftext|>406 Not Acceptable⏎HTTP response status code 406 Not Acceptable is a client error that is returned by the server to indicate that

  2. #2 peak

    used because the server supplies a default representation instead. The rationale is that it assumes that the client will prefer something over nothing. However, when the 406 Not Acceptable response is returned, the message body will contain a list of available

  3. #3 peak

    in a circle⏎(C) 2009 Copyright Math Open Reference. All rights reserved<|endoftext|>Use chain rule⏎F'(x)

  4. #4 peak

    the search results.⏎In the example, the client requests a resource in the French language but the server responds with the 406 Not Acceptable response because only German and English versions of the document are available.⏎

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. L3:2599 weakly suppresses
  2. L1:2115 weakly suppresses
  3. L4:739 weakly suppresses
  4. L5:198 weakly excites unclear: weak "and" in lists
  5. L3:1770 weakly excites
  6. L2:1073 weakly excites
  7. L1:2915 weakly excites
  8. L2:2615 weakly suppresses
  9. L6:43 weakly suppresses
  10. L2:686 weakly excites
  11. L2:2756 weakly excites
  12. L6:1608 weakly suppresses
  13. L0:807 weakly suppresses
  14. L6:704 weakly excites
  15. L6:1430 weakly excites
  16. L2:1827 weakly excites
  17. L7:21 weakly excites
  18. L3:733 weakly excites
  19. L0:2385 weakly suppresses
  20. L6:686 weakly suppresses
  21. L4:1002 weakly excites
  22. L6:29 weakly suppresses
  23. L1:1099 weakly excites
  24. L7:1985 weakly suppresses

Feeds later layers

  1. L9:283 excites
  2. L10:217 excites
  3. L9:250 excites
  4. L16:656 excites
  5. L11:567 excites
  6. L15:181 suppresses
  7. L10:1082 excites
  8. L9:84 suppresses
  9. L12:959 excites
  10. L12:1820 excites
  11. L9:1396 suppresses
  12. L13:371 suppresses
  13. L17:806 suppresses
  14. L9:270 excites
  15. L13:788 excites
  16. L15:264 excites
  17. L12:168 excites
  18. L9:597 excites
  19. L12:74 excites
  20. L9:1607 excites
  21. L19:1130 excites
  22. L10:457 weakly excites
  23. L10:121 weakly suppresses
  24. L13:1550 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 -0.68 and -0.623 values between -0.62 and -0.567 values between -0.56 and -0.5012 values between -0.50 and -0.4332 values between -0.43 and -0.3776 values between -0.37 and -0.31102 values between -0.31 and -0.25212 values between -0.25 and -0.18377 values between -0.18 and -0.12547 values between -0.12 and -0.06711 values between -0.06 and 0.00640 values between 0.00 and 0.07509 values between 0.07 and 0.13311 values between 0.13 and 0.19171 values between 0.19 and 0.25114 values between 0.25 and 0.3238 values between 0.32 and 0.3822 values between 0.38 and 0.4413 values between 0.44 and 0.506 values between 0.50 and 0.573 values between 0.57 and 0.631 values between 1.75 and 1.82offnegativepositiveoften

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. игра
  2. 骨
  3. иг
  4. bcrypt
  5. IFICATIONS
  6. agues
  7. фигур
  8. resher
  9. fenced
  10. bcrypt

Suppresses

  1. otle
  2. pil
  3. gelişme
  4. sol
  5. 指
  6. Index
  7. mittel
  8. Phần
  9. struction
  10. hospitality