Neuron Explorer

Qwen3-0.6B

unclear: document ends and list breaks

A neuron in layer 14 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron fires at unusual breaks in text, such as the end of a document, a number split by a symbol, or the start of a new clause. It is hard to say more than that, and turning it up made the model repeat itself.

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

    , towards the location of the monument. Where it is not available, the satellite view is shown instead.⏎Latitude: 52.5162 / 52°30'58"N⏎Longitude: -3.0

  2. #2 peak

    Pradesh, it is more comfortable to wear it over a shirt than as a single⏎piece of clothing.⏎Q4. What do you wear on veshti?⏎<|endoftext|>Copyright (c) Arvin S. Quist⏎INTRODUCTION TO CLASSIFICATION

  3. #3 peak

    Erica plukenetii, E. lineata and E breviflora. They have been sunk into E

  4. #4 peak

    Check out the full range of plans here: https://www.edplace.com/blog/tag/learning_plans.⏎What are the EdPlace learning plans?⏎Put simply, our teachers have created a half-termly (or 6 week) learning plan

  5. #5 peak

    bacteria bind to the sugar chain of UMOD and are eventually removed through the urine. During recent years, evidence has shown that glycoprotein 2 (GP2), a UMOD-like molecule produced in the pancreas

  6. #6 peak

    Studies using Geophysical techniques⏎- Implement and Development of Geophysical Inversion Methods⏎- Petroleum and Mineral Exploration⏎- Physics of the Earth and other Planets<|endoftext|>

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: I am trying to find the number of the number of the number of the number of the number of the number of the number of the number of the number of the number of... the number of... the number of... the number of...

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

    creates metafiction.⏎One example of metafiction can be found on the very first page. Ondaatje transitions from writing in third person, using pronouns like "her" and "he," to writing in first person. True, it

  2. #2 peak

    first naval battle in which victory was decided by aircraft alone, Dauntless dive-bombers fought well alongside other US aircraft. Each side lost one carrier (the Japanese carrier Shoho being sunk by Dauntless and

  3. #3 peak

    82 until Zelaya’s ouster in June, the Honduran government appeared to be relatively stable. The country’s two major political parties, the moderate Liberal Party (of which Zelaya is a member) and the more

  4. #4 peak

    5 to 2008.⏎As a child in Cleveland, Ohio McKinney was shaped by his father who challenged racism, and invited civil rights leaders such as Thurgood Marshall to speak frequently at his church. McKinney attended Morehouse

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. L12:1640 excites
  2. L7:307 excites
  3. L13:732 excites
  4. L13:1897 suppresses
  5. L9:557 suppresses
  6. L13:2181 excites
  7. L13:1314 excites
  8. L13:1483 weakly excites
  9. L8:481 weakly suppresses
  10. L13:700 weakly excites
  11. L13:1333 weakly excites
  12. L0:1637 weakly suppresses
  13. L11:3062 weakly suppresses
  14. L10:1459 weakly suppresses
  15. L10:681 weakly excites
  16. L12:2366 weakly suppresses
  17. L12:810 weakly excites
  18. L13:1376 weakly excites
  19. L7:231 weakly excites
  20. L1:136 weakly excites
  21. L9:1414 weakly suppresses opening parenthesis after copyright or year
  22. L13:648 weakly suppresses
  23. L10:1090 weakly excites
  24. L13:505 weakly suppresses

Feeds later layers

  1. L16:431 suppresses
  2. L15:2967 excites
  3. L16:413 suppresses
  4. L16:128 suppresses inside numbers and citations
  5. L16:468 suppresses
  6. L15:1653 excites
  7. L16:267 suppresses
  8. L16:737 weakly suppresses
  9. L15:1452 weakly excites
  10. L16:2161 weakly excites
  11. L16:860 weakly excites
  12. L16:648 weakly suppresses
  13. L23:2096 weakly excites
  14. L15:2616 weakly excites
  15. L16:2700 weakly excites
  16. L15:131 weakly suppresses
  17. L16:1906 weakly excites
  18. L16:349 weakly suppresses
  19. L18:1009 weakly suppresses
  20. L17:503 weakly suppresses
  21. L19:347 weakly excites
  22. L15:724 weakly suppresses
  23. L25:1403 weakly excites
  24. L16:665 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 -6.34 and -6.121 values between -3.75 and -3.541 values between -3.54 and -3.321 values between -3.32 and -3.103 values between -3.10 and -2.891 values between -2.89 and -2.672 values between -2.67 and -2.462 values between -2.46 and -2.247 values between -2.24 and -2.021 values between -2.02 and -1.816 values between -1.81 and -1.595 values between -1.59 and -1.388 values between -1.38 and -1.1612 values between -1.16 and -0.9529 values between -0.95 and -0.7345 values between -0.73 and -0.51100 values between -0.51 and -0.30394 values between -0.30 and -0.082297 values between -0.08 and 0.13579 values between 0.13 and 0.35229 values between 0.35 and 0.5789 values between 0.57 and 0.7851 values between 0.78 and 1.0012 values between 1.00 and 1.2114 values between 1.21 and 1.4313 values between 1.43 and 1.644 values between 1.64 and 1.861 values between 2.08 and 2.29offnegativepositiveoften

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. 欺
  2. リスク
  3. allocation
  4. lover
  5. (New
  6. ontology
  7. ||(
  8. 莫斯
  9. �建
  10. 未来发展

Suppresses

  1. śni
  2. uble
  3. among
  4. uci
  5. 在中国
  6. urch
  7. صة
  8. erra
  9. 特朗
  10. ebra