Neuron Explorer

Qwen3-0.6B

period ending an opening topic sentence

A neuron in layer 11 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

It fires on the full stop that closes the first sentence of a web article paragraph, the sentence that announces what the paragraph will discuss. Turning it up made the model ramble and repeat words.

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

    , saving energy and reducing utility costs.⏎Choosing the Right Replacement Parts⏎When it comes to repairing or upgrading an HVAC system, choosing the correct replacement parts is critical. Quality parts ensure compatibility with the system and extend its lifespan. For example, using

  2. #2 peak

    & remix collections to collaborate.⏎Organize your curriculum with collections. Easy!⏎Have time to be more creative & energetic with your students!⏎Students explore the Orange Revolution. In this Orange Revolution lesson, students examine the climate of the Ukrainian protest and investigate

  3. #3 peak

    with...⏎10th - 12th English Language Arts⏎Lesson Plan: The State of the Union⏎Learners explore the state of the union. In this U.S. government lesson, students analyze the 1998

  4. #4 peak

    trusts.⏎How and When Does a Living Trust Terminate?⏎The manner in which a trust terminates depends, to some extent, on the type of trust in question. If the trust is a revocable living trust or a testamentary trust,

  5. #5 peak

    coding skills have become invaluable for developers looking to make a mark in the industry.⏎The Challenges of Coding for Apps⏎Coding for apps presents its fair share of challenges. Developers must understand the unique requirements of the

  6. #6 peak

    2, 2021, The Florida Fish and Wildlife Conservation Commission (FWC) met to discuss a myriad of topics that affect the South Florida area. Among the points of discussion, a major proposal made the agenda – to lift the

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: "The New and Beautiful" and the "Fruit of the Bean Tree." which of the most important and enduring consequences of human Value and Intelligence, Dimensions, Full Human Dimensions, and Series of Dimensions, Perhaps We Ever Betterment of All

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

    tetracycline resistance was maintained in spite of the declining concentration of tetracycline. (C) 2002 Elsevier Science B.V. All rights reserved.⏎|Publication status||Published - 2002|⏎

  2. #2 peak

    information.<|endoftext|>História (São Paulo)⏎On-line version ISSN 1980-4369⏎DOMINGUES, Beatriz Helena. The guarani jesuitic missions in the context of Enlightenment. História [

  3. #3 peak

    to those before you that:-" If you join gods with Allah, your deeds shall be fruitless, and you surely will be one of those who lose. (all spiritual goodnesses)⏎THE HOLY QURAN - S39:

  4. #4 peak

    to it (The surface tool only works on a single object). Now go into segment select mode and select all those extra loose segmenets and delete them. Hmm it looks like a rounded pipe to me . Lets<|endoftext|>Jonathan D. Jaip

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. L10:478 suppresses
  2. L9:2302 excites
  3. L9:16 suppresses unclear: periods inside numbers and names
  4. L8:541 excites
  5. L9:2447 excites
  6. L10:1580 suppresses
  7. L5:155 suppresses
  8. L6:2302 suppresses 'and' joining two parallel items
  9. L4:7 excites
  10. L5:168 suppresses
  11. L6:250 excites
  12. L4:848 excites
  13. L10:612 excites full stop ending boilerplate or a title
  14. L5:1653 excites
  15. L9:1802 excites
  16. L3:259 weakly suppresses
  17. L8:361 weakly excites
  18. L3:671 weakly suppresses
  19. L0:476 weakly excites
  20. L2:9 weakly suppresses sentence-final periods vs dots in code
  21. L9:458 weakly excites
  22. L6:305 weakly suppresses
  23. L8:740 weakly excites
  24. L6:2621 weakly suppresses

Feeds later layers

  1. L17:347 excites full stops ending informative sentences
  2. L14:155 excites sentence-ending periods in prose
  3. L20:940 excites
  4. L16:1545 excites
  5. L18:791 excites
  6. L17:1105 excites
  7. L19:344 excites periods ending explanations
  8. L13:9 suppresses
  9. L17:1772 excites
  10. L12:1062 excites
  11. L16:432 excites period after month or initial
  12. L20:473 excites
  13. L20:962 excites decimal points and dots in numbers
  14. L20:505 excites
  15. L19:2097 excites sentence-end punctuation in speech
  16. L12:1837 excites
  17. L16:2535 excites
  18. L13:53 excites period ending a claim before elaboration
  19. L21:1086 excites end of sentence before a claim or statistic
  20. L19:94 suppresses
  21. L17:990 excites
  22. L19:384 excites
  23. L12:2141 suppresses
  24. L27:145 excites dots inside web addresses

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.

7 values between -1.11 and -0.9381 values between -0.93 and -0.75376 values between -0.75 and -0.581128 values between -0.58 and -0.401466 values between -0.40 and -0.22547 values between -0.22 and -0.04126 values between -0.04 and 0.1432 values between 0.14 and 0.3220 values between 0.32 and 0.5010 values between 0.50 and 0.6815 values between 0.68 and 0.866 values between 0.86 and 1.046 values between 1.04 and 1.227 values between 1.22 and 1.407 values between 1.40 and 1.587 values between 1.58 and 1.765 values between 1.76 and 1.946 values between 1.94 and 2.123 values between 2.12 and 2.305 values between 2.30 and 2.486 values between 2.48 and 2.654 values between 2.65 and 2.834 values between 2.83 and 3.015 values between 3.01 and 3.192 values between 3.19 and 3.374 values between 3.37 and 3.555 values between 3.55 and 3.736 values between 3.73 and 3.912 values between 3.91 and 4.093 values between 4.09 and 4.271 values between 4.27 and 4.451 values between 4.45 and 4.631 values between 4.63 and 4.812 values between 4.99 and 5.171 values between 5.35 and 5.521 values between 5.88 and 6.06offnegativepositiveoften

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. tering
  2. 目的是
  3. tracted
  4. 届时
  5. 不过
  6. 一方面
  7. 家都知道
  8. これ
  9. 回首
  10. 然而

Suppresses

  1. nostic
  2. y
  3. .&
  4. ィ
  5. chlor
  6. punto
  7. 万户
  8. olg
  9. horn
  10. 西部