Neuron Explorer

Qwen3-0.6B

"and" joining two parts of a name or phrase

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 word "and" when it links two parts of a title, name or description, such as "professor of global and environmental health". Turned up, the model got stuck repeating "I can't" or numbers, so the steering shows no clear theme.

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

    as diabetes and obesity are closely related, our findings provide clear evidence that chronic exposure to air pollution increases the risk for developing obesity," said Duke University professor of global and environmental health, Junfeng Zhang. “If translated and verified in humans,

  2. #2 peak

    about WWII through the story of Anne Frank.⏎The back cover of the book carries an endorsement by the popular children’s author Jacqueline Wilson: “Exquisitely and imaginatively illustrated, this is a sensitive retelling of the moving story of Anne

  3. #3 peak

    ?⏎In simple terms, cloud computing is a technology used for storing and accessing data, application and programs over the Internet. Cloud computing refers to the delivery of hosted and computing services including servers, storage, databases, networking, analytics, and intelligence over

  4. #4 peak

    do volunteer work during the course. Prerequisite: permission of instructor.⏎Course Description and Rationale:⏎A 2006 report by the Corporation for National and Community Service found that 3.3 million college students volunteered in 20

  5. #5 peak

    environment, and this work opens the door to new biological experiments," said Thomas V. O'Halloran, the study's senior author and the Charles E. and Emma H. Morrison Professor of Chemistry in the Weinberg College of Arts and Sciences

  6. #6 peak

    Harper (American, 1922–2007). Kids of all ages can enjoy playing the game and learning from the accompanying booklet about familiar and exotic creatures.⏎Sample booklet text:⏎Black bears usually live in forests and can climb

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 want to create a space for a research project in a remote research university in the Pacific Islands, research university in the Pacific Islands, research university in the research university in the research university in the research university in the research university in the

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

    that quantity. It states that the amount of the conserved quantity at a point or within a volume can only change by the amount of the quantity which flows in or out of the

  2. #2 peak

    redundancy, so if one drive breaks down, the data will be accessible from the remaining ones, and the second is improved performance as the input/output, or reading/writing operations will be distributed among different drives. There are different RAID types depending on

  3. #3 peak

    an interaction among 3 basic clusters of human traits:⏎• Above-average general &/or specific abilities⏎• High levels of task commitment (motivation)⏎• High

  4. #4 peak

    emia||1%||Ruled out|⏎How often do you get bacterial, viral, or fungal infections? For example, skin, ear, gastrointestinal, chest/respiratory, mouth, genitourinary. Symptoms usually include fever, pain

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. L14:801 excites 'and' or 'or' joining list items
  2. L9:92 suppresses "and" or "or" joining two items
  3. L14:2276 excites opening bracket after a term or number
  4. L14:2823 excites
  5. L9:1820 excites
  6. L8:70 excites 'and' between stacked superlatives
  7. L9:2010 suppresses
  8. L14:2202 excites
  9. L13:131 excites 'and' joining parallel phrases
  10. L11:406 suppresses
  11. L10:2046 suppresses
  12. L14:706 excites
  13. L14:2627 excites 'or' or 'and' joining two items
  14. L14:1485 excites
  15. L14:745 suppresses
  16. L12:1786 excites
  17. L10:74 excites
  18. L12:2548 excites
  19. L9:1899 suppresses
  20. L4:57 suppresses
  21. L13:2356 excites
  22. L12:51 excites
  23. L6:427 suppresses
  24. L14:2865 suppresses

Feeds later layers

  1. L16:2580 excites 'and' linking food or health nouns
  2. L17:2274 excites
  3. L18:1625 excites
  4. L16:1937 excites 'and' joining two time or quantity ranges
  5. L18:1572 excites
  6. L18:1270 weakly excites
  7. L18:2354 weakly excites
  8. L18:2924 weakly excites
  9. L17:3025 weakly excites
  10. L23:1446 weakly excites
  11. L17:1802 weakly excites
  12. L20:2794 weakly excites
  13. L17:595 weakly excites
  14. L16:1462 weakly excites
  15. L21:409 weakly excites
  16. L21:197 weakly suppresses
  17. L18:1272 weakly suppresses
  18. L24:87 weakly excites
  19. L24:1294 weakly excites
  20. L19:1823 weakly excites 'and' inside organisation or paired names
  21. L17:48 weakly suppresses
  22. L24:1419 weakly excites
  23. L21:2225 weakly excites
  24. L18:1857 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 -7.83 and -7.441 values between -5.49 and -5.101 values between -5.10 and -4.711 values between -4.71 and -4.321 values between -4.32 and -3.932 values between -3.93 and -3.545 values between -3.54 and -3.153 values between -3.15 and -2.764 values between -2.76 and -2.3712 values between -2.37 and -1.987 values between -1.98 and -1.5925 values between -1.59 and -1.2023 values between -1.20 and -0.81146 values between -0.81 and -0.422175 values between -0.42 and -0.031396 values between -0.03 and 0.3647 values between 0.36 and 0.7518 values between 0.75 and 1.1411 values between 1.14 and 1.539 values between 1.53 and 1.925 values between 1.92 and 2.313 values between 2.31 and 2.703 values between 2.70 and 3.091 values between 3.09 and 3.482 values between 3.48 and 3.873 values between 3.87 and 4.261 values between 4.65 and 5.041 values between 6.60 and 6.991 values between 7.38 and 7.77offnegativepositiveoften

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. نك
  2. _hpp
  3. raries
  4. acies
  5. äft
  6. 況
  7. bish
  8. Addresses
  9. addresses
  10. 矧

Suppresses

  1. own
  2. Returns
  3. oge
  4. (<
  5. dae
  6. 喃
  7. licht
  8. dee
  9. льн
  10. 父子