Neuron Explorer

Qwen3-0.6B

'and' or 'or' joining the last list items

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 to the word 'and' when it joins the final two items in a list, such as 'signs and symptoms' or 'wings; or ground'. When it was turned down, the model kept writing long lists joined by 'or', which suggests it helps track list-style joining 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

    latter is preferable as you can plant these in the ground and they will rot away)⏎- plastic containers – these also are takeout containers from a deli and they have the bonus of a domed clear plastic lid

  2. #2 peak

    USDA and U.S. Food and Drug Administration, raw chicken (regardless of if it’s whole; in pieces such as breasts, thighs, drumsticks, and wings; or ground) should be stored

  3. #3 peak

    to see a doctor⏎Call your doctor if you or your child has any of these signs and symptoms:⏎- A sore throat accompanied by tender, swollen lymph glands⏎- A

  4. #4 peak

    . For example: add wheels, doors, windows, lights to the car; or add details to the scene like a road, traffic light, stop sign, and trees.<|endoftext|>Coral Reefs are the “Rainforests” of the

  5. #5 peak

    of injury in a fire or other accident.⏎- Install aftermarket parts correctly. Incorrectly-installed cruise control, stereos, and other aftermarket parts may pose electrical risks that can start a fire. Make sure these

  6. #6 peak

    a report is a document that presents or communicates to a target audience the findings (results, guidelines or outcomes) of technical work that was carried out by an individual and/or group. When this definition is carefully looked at, it becomes evident that it

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: or the specific, or the specific, or the specific, or the specific, or their specific, or their own, or something, or, or, or, or, or, or, or, or, or, or, or

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

    paper, or report: you'll put in too much or too little information. Don't rely on the way material was phrased in the article, paper, or report: summarize information in a new way. Don'ts Do not commence

  2. #2 peak

    arest definitions of Cap and Trade: A cap and trade system is a method for managing pollution, with the end goal of reducing the overall pollution in a nation, region, or industry. Many proponents of pollution control support the concept of cap and trade

  3. #3 peak

    you⏎determine that num is not divisible by 5 and 7, then increase⏎count by one.⏎Share your code⏎Show a friend, family member, or teacher what you've done!<|endoftext|>cigarette butts are the

  4. #4 peak

    divisions among republicans, Socialists, and leftists had permitted conservatives to come<|endoftext|>More people are donating one of their kidneys to a loved one, a friend, even a stranger, and now a move is on to make sure those donors really

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. L14:1107 strongly suppresses
  2. L14:1174 strongly excites
  3. L14:1351 strongly excites
  4. L14:1834 excites
  5. L14:2615 suppresses
  6. L14:516 excites
  7. L14:346 excites
  8. L14:3034 excites
  9. L14:2986 excites
  10. L14:1500 suppresses
  11. L14:1571 suppresses
  12. L13:2939 excites
  13. L14:2835 suppresses
  14. L13:569 suppresses
  15. L12:660 suppresses
  16. L12:2393 weakly excites
  17. L11:119 weakly suppresses
  18. L13:2826 weakly suppresses
  19. L14:589 weakly suppresses
  20. L13:562 weakly suppresses
  21. L10:967 weakly excites
  22. L14:3046 weakly excites
  23. L14:1231 weakly suppresses
  24. L1:1925 weakly suppresses

Feeds later layers

  1. L25:41 excites
  2. L19:289 excites
  3. L23:223 suppresses
  4. L25:1270 suppresses
  5. L20:1334 suppresses
  6. L18:2197 suppresses
  7. L27:376 suppresses
  8. L18:2362 suppresses
  9. L17:2166 suppresses
  10. L16:2289 suppresses 'or' linking two alternatives
  11. L17:1897 suppresses
  12. L26:2473 suppresses
  13. L18:1786 suppresses
  14. L20:1599 suppresses
  15. L20:2307 suppresses "some" in "some kind" and "at some point"
  16. L21:2124 suppresses
  17. L17:2671 suppresses the word matter in no matter phrases
  18. L18:2896 suppresses
  19. L26:347 suppresses the 're' in self-reliant and similar
  20. L25:1394 suppresses
  21. L18:1138 weakly suppresses
  22. L26:779 weakly suppresses
  23. L16:935 weakly excites
  24. L16:1388 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 -6.48 and -6.132 values between -3.65 and -3.302 values between -2.23 and -1.883 values between -1.88 and -1.533 values between -1.53 and -1.1715 values between -1.17 and -0.8261 values between -0.82 and -0.46609 values between -0.46 and -0.112791 values between -0.11 and 0.24381 values between 0.24 and 0.6013 values between 0.60 and 0.9511 values between 0.95 and 1.313 values between 1.31 and 1.664 values between 1.66 and 2.011 values between 2.01 and 2.371 values between 2.72 and 3.082 values between 3.08 and 3.431 values between 3.43 and 3.781 values between 4.14 and 4.492 values between 4.49 and 4.851 values between 7.32 and 7.68offnegativepositiveoften

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. &r
  2. -
  3. oooo
  4. Orient
  5. 真实性
  6. %
  7. 上下
  8. u
  9. _PACK
  10. 上

Suppresses

  1. 或多
  2. 或是
  3. 或者
  4. 或者是
  5. 或者其他
  6. 或其他
  7. or
  8. 或
  9. 或
  10. atau