Neuron Explorer

Qwen3-0.6B

"and" joining paired verbs

A neuron in layer 15 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 responds weakly to the word "and" when it links two actions or qualities, like "preserving and protecting" or "analyze, synthesize, and evaluate". Its activations are small compared with the other neurons here. Turning it up did not give a clear theme in the model's writing.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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

    Act no. 17 of 1996, establishes that “every person in Zambia has been and shall continue to be entitled to the fundamental rights and freedoms of the individual

  2. #2 peak

    of life, and our country's future.⏎Why It's Important⏎As stewards of the environment, we are responsible for preserving and protecting our resources for ourselves and for future generations.⏎Getting Back To Basics⏎Rec

  3. #3 peak

    engage and impact their community, both locally and globally. In all aspects of their lives, UMass Dartmouth graduates skillfully locate, analyze, synthesize, and evaluate information and can communicate their knowledge effectively and creatively. They have depth in a

  4. #4 peak

    ?⏎Since ancient times, its influence has permeated East Asian culture, and as it spread to different countries and regions, the practice of Taoism became increasingly diverse and localized to some extent. However, the belief in immortality is the common characteristic

  5. #5 peak

    Dashes: Adventures at the Dawn of the Wireless Age" is at http://vubooks.com<|endoftext|>Child health nursing is designed to help the students to develop and understand the concept of child health, the trends in child care and health problems of

  6. #6 peak

    , random.choice (fruit)⏎Any Tutorial Videos?<|endoftext|>What is globalisation?⏎Globalisation is the process by which the world is becoming increasingly interconnected as a result of massively increased trade and cultural exchange. Globalisation has increased the production

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 and and. and and. and. and. and. and. and. and. and. and. and. and. and. and. and. and. and. and. and. and. and. and. and

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

    They washed after rising and⏎both before and after the main meals. As mouth wash they used another salt solution<|endoftext|>About Apollo Missions"That’s one small step for a man, one giant leap for mankind...” nnOn 20 July

  2. #2 peak

    At that point, the Bible tells us Jesus was deeply moved and wept.⏎Jesus went to the cave where Lazarus was buried and shouted, “Lazarus

  3. #3 peak

    to help create an elementary version of the highly-successful program.⏎Additional Meeting Notes⏎There are a limited number of speaking slots. It is a first come, first served system.<|endoftext|>Fast Food Near Urban Schools Causes Obesity In Minority Children?⏎Does

  4. #4 peak

    this article. Of course, they can.⏎You’ll surely learn something from these moral stories at the end.⏎Read some other English moral stories.⏎Hence without wasting your valuable time, let’s start writing some of the best moral stories in English

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. L10:873 excites
  2. L14:1542 suppresses
  3. L11:223 excites
  4. L10:556 excites
  5. L11:1286 weakly excites
  6. L12:1027 weakly excites
  7. L11:840 weakly suppresses
  8. L10:1278 weakly suppresses
  9. L10:370 weakly suppresses
  10. L14:1319 weakly excites
  11. L11:1826 weakly excites
  12. L2:24 weakly suppresses
  13. L4:267 weakly excites
  14. L4:92 weakly suppresses
  15. L7:72 weakly excites
  16. L4:1050 weakly suppresses
  17. L9:530 weakly suppresses
  18. L13:2018 weakly suppresses
  19. L3:245 weakly suppresses
  20. L10:313 weakly excites
  21. L14:731 weakly suppresses
  22. L6:701 weakly suppresses
  23. L14:0 weakly suppresses
  24. L10:1997 weakly excites

Feeds later layers

  1. L16:2306 excites
  2. L21:3002 weakly suppresses
  3. L16:2927 weakly excites
  4. L25:2355 weakly suppresses
  5. L25:1175 weakly suppresses
  6. L19:659 weakly suppresses
  7. L24:1802 weakly suppresses
  8. L19:497 weakly excites
  9. L18:2041 weakly excites
  10. L18:1009 weakly excites
  11. L23:1302 weakly suppresses
  12. L18:2701 weakly suppresses
  13. L18:1413 weakly suppresses
  14. L17:1781 weakly suppresses
  15. L27:218 weakly excites
  16. L16:2860 weakly excites
  17. L23:728 weakly suppresses
  18. L18:701 weakly suppresses
  19. L17:2172 weakly excites
  20. L16:2856 weakly excites
  21. L18:2433 weakly suppresses
  22. L24:227 weakly excites
  23. L22:994 weakly suppresses
  24. L21:109 weakly suppresses

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 -1.05 and -0.962 values between -0.67 and -0.573 values between -0.57 and -0.486 values between -0.48 and -0.3840 values between -0.38 and -0.29147 values between -0.29 and -0.19383 values between -0.19 and -0.101069 values between -0.10 and -0.001047 values between -0.00 and 0.10408 values between 0.10 and 0.19221 values between 0.19 and 0.29147 values between 0.29 and 0.38108 values between 0.38 and 0.4870 values between 0.48 and 0.5758 values between 0.57 and 0.6739 values between 0.67 and 0.7634 values between 0.76 and 0.8626 values between 0.86 and 0.9517 values between 0.95 and 1.0522 values between 1.05 and 1.1511 values between 1.15 and 1.2410 values between 1.24 and 1.346 values between 1.34 and 1.437 values between 1.43 and 1.539 values between 1.53 and 1.623 values between 1.62 and 1.723 values between 1.81 and 1.913 values between 1.91 and 2.001 values between 2.00 and 2.101 values between 2.10 and 2.201 values between 2.20 and 2.291 values between 2.29 and 2.392 values between 2.39 and 2.481 values between 2.58 and 2.671 values between 2.67 and 2.77offnegativepositiveoften

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. REE
  2. IFA
  3. (hero
  4. uesday
  5. arya
  6. ULA
  7. ceptive
  8. Bureau
  9. agina
  10. specifically

Suppresses

  1. stype
  2. 内置
  3. 虬
  4. 斯顿
  5. erset
  6. intel
  7. lems
  8. esda
  9. émon
  10. fills