Neuron Explorer

Qwen3-0.6B

end of a web document

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 reacts to the spot where a web page or document ends and an unrelated one could begin, marked by the end of text token. Turning it down made the model ramble and repeat itself, and turning it up gave off-topic text.

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

    |⏎|Required to build||libstdc++|⏎|Required to build||fortran|⏎This entry (in part or in whole) was last reviewed on 4 February 2012.⏎Permission is granted to copy, distribute

  2. #2 peak

    speech; the second deals with the relative clause and sentence combining by addition; and the third section covers the verbals.⏎Includes a test/answer key booklet.<|endoftext|>extinction coefficient - (Mar/17/2006 )⏎how

  3. #3 peak

    Education.⏎A minimum resolution of 1024 x 768 is recommended.⏎To access this document off-line, you can print this document.<|endoftext|>Maybe you’ve had raw milk, seen it for sale in your local health food

  4. #4 peak

    of tissue are taken for further analysis. It may also be necessary to retain entire organs for a more detailed examination. This is strictly limited to cases where such retention<|endoftext|>Published August 30, 2004⏎by Diane Pub Co

  5. #5 peak

    20 years, and the best bird of the count has been identified on LyondellBasell property several years.”⏎Michael Kennedy worked for LyondellBas<|endoftext|>Discover the cosmos! Each day a different image or photograph of our fascinating universe is

  6. #6 peak

    to the youngest visitors to the museum. They can practice the alphabet as they travel from letter to letter in the maze. An open area at the front with toddler<|endoftext|>Mental Health and Wellbeing for Students⏎Moat gives students a platform in

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 number, 331, is the number that is in the day time is in the time of "the time is 331 on the time, that's a person is time, is, which is the time,

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

    ation/the-omnivores-delusion-against-the-agri-intellectuals/⏎Leave a ReplyWant to join the discussion?⏎Feel free to contribute!<|endoftext|>Tree Roots Can Wreak Havoc on Pipes — How to Best Prevent Pipe

  2. #2 peak

    trip the circuit breaker.⏎Email Based, Online Homework Assignment Help in Classification Of Relays⏎Transtutors is the best place to get answers to all your doubts regarding classification of relays with examples. Transtutors has a vast panel of experienced

  3. #3 peak

    , title, abstract, bibliographic or download information, contact: (Michael McNulty)⏎If references are entirely missing, you can add them using this form.<|endoftext|>Farmers fight drought with move to digital age⏎21 August 20

  4. #4 peak

    These include physical features of the<|endoftext|>Learn vocabulary, terms, and more with flashcards, games, and other study tools. Optional activities are designed to enhance understanding andor to provide additional practice. Each

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. L13:2359 excites
  2. L13:1084 suppresses
  3. L14:1736 excites
  4. L12:1648 suppresses
  5. L13:2354 excites
  6. L13:3045 suppresses
  7. L12:2466 excites
  8. L13:2027 suppresses
  9. L14:2164 suppresses
  10. L12:2323 excites
  11. L11:122 suppresses
  12. L14:80 suppresses
  13. L12:2873 excites
  14. L13:2946 excites
  15. L13:2147 suppresses
  16. L12:2943 excites
  17. L14:1042 suppresses
  18. L12:2684 excites
  19. L14:1641 suppresses
  20. L14:2941 excites
  21. L13:2014 suppresses
  22. L9:2085 suppresses
  23. L11:1675 suppresses
  24. L12:2668 suppresses

Feeds later layers

  1. L16:248 suppresses
  2. L17:892 excites
  3. L16:2477 suppresses
  4. L16:200 suppresses
  5. L16:1543 excites
  6. L16:1262 suppresses
  7. L20:543 excites
  8. L27:828 excites
  9. L19:929 excites
  10. L22:157 excites
  11. L18:129 suppresses
  12. L18:45 suppresses
  13. L19:229 suppresses
  14. L21:105 excites
  15. L23:206 excites
  16. L20:312 excites
  17. L27:609 excites
  18. L27:856 excites
  19. L20:33 suppresses
  20. L24:226 excites
  21. L16:2151 weakly excites
  22. L21:12 weakly suppresses
  23. L19:251 weakly excites
  24. L27:581 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.

2 values between -1.20 and -1.082 values between -0.85 and -0.732 values between -0.73 and -0.6213 values between -0.50 and -0.3864 values between -0.38 and -0.27314 values between -0.27 and -0.151081 values between -0.15 and -0.041624 values between -0.04 and 0.08624 values between 0.08 and 0.20122 values between 0.20 and 0.3129 values between 0.31 and 0.439 values between 0.43 and 0.541 values between 0.54 and 0.667 values between 0.66 and 0.781 values between 0.78 and 0.891 values between 0.89 and 1.013 values between 1.01 and 1.121 values between 1.12 and 1.241 values between 2.05 and 2.171 values between 2.17 and 2.283 values between 2.52 and 2.631 values between 2.86 and 2.981 values between 2.98 and 3.101 values between 3.33 and 3.44offnegativepositiveoften

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. :",⏎
  2. *)⏎⏎
  3. )")⏎⏎
  4. ++)⏎
  5. /"⏎⏎
  6. *}⏎⏎
  7. --⏎⏎
  8. …”⏎⏎
  9. cual
  10. .";⏎⏎

Suppresses

  1. D
  2. D
  3. O
  4. l
  5. d
  6. atory
  7. L
  8. N
  9. ,