Neuron Explorer

Qwen3-0.6B

'and' before a second verb phrase

A neuron in layer 17 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 fires on "and" when it links a finished phrase to a new verb, as in "X is measured, and is...". When it was pushed down, the model repeated the word "because", and when pushed up it produced lists of "and".

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

    is to drive with a BAC of .08 percent.⏎Blood alcohol concentration, or BAC, describes the amount of alcohol contained in an individual’s blood and is measured as weight per unit of volume. The measurement is usually

  2. #2 peak

    but your systolic pressure has lowered. This condition is referred to as systolic hypotension and is mainly defined by signs and symptoms, rather than a specific blood pressure reading.⏎

  3. #3 peak

    ís Treaty between the United States and Spain was negotiated by Secretary of State John Quincy Adams and the Spanish Minister to the United States, Don Luis de Onís, and signed in February 1819. The principal elements in the treaty were

  4. #4 peak

    shoulder injuries. While physiotherapy is recommended in all of these cases, other types of rehabilitation may be needed as well. Neurological disorders involve the nervous system and may cause paralysis to a part of the body. Examples are stroke, Alzheimer’s

  5. #5 peak

    if you are making plans.⏎There are spring parades that occur some three weeks before Easter. These are called Sommertagszug (Summer Day Parade) and are basically a time when people tell winter that it’s time to leave.⏎Also

  6. #6 peak

    used for a myriad of symptoms affecting your ability to think. Brain fog can include things like memory problems, lack of mental clarity, and an inability to focus, and put thoughts into words. Some people describe it as mental fatigue.⏎You know –

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: because that because because because because because because because because because because because because because because because because because because because because because because because because because because because because because because because because because because because because because because because because because because because because because because

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

    be auctioned, given up to an animal rescue group, or put to sleep.⏎Compared to other states’ cruelty laws, Texas animal cruelty laws are unique in a couple of ways. First, Texas criminal cruelty laws only apply to domesticated

  2. #2 peak

    man of the people.” It’s even odder that he’s accused of being a “leftist.” In fact, he’s

  3. #3 peak

    ’ stumbled upon one of the world’s oldest wrecks. At nearly 3,000 years old, this Bronze Age find is particularly significant, not just because it predates the birth of Christ by a millennium, but because in addition

  4. #4 peak

    Anopheles females lay about 50 to 200 eggs. Eggs are laid directly on water and are unique in having floats on either side. These eggs are not resistant to drying and hatch within

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. L16:923 excites
  2. L15:1007 suppresses
  3. L3:1055 suppresses
  4. L6:1461 weakly excites
  5. L16:1429 weakly excites
  6. L9:954 weakly suppresses
  7. L14:2638 weakly suppresses
  8. L1:1763 weakly excites
  9. L12:320 weakly suppresses
  10. L11:1156 weakly suppresses
  11. L4:2224 weakly suppresses
  12. L11:158 weakly suppresses
  13. L7:469 weakly excites
  14. L13:789 weakly suppresses
  15. L16:2512 weakly suppresses
  16. L1:784 weakly excites
  17. L13:797 weakly excites
  18. L10:1338 weakly suppresses digits inside numbers
  19. L15:788 weakly excites
  20. L9:2318 weakly excites 'of' in 'School of' institution names
  21. L1:1741 weakly suppresses
  22. L14:1748 weakly excites
  23. L10:1002 weakly excites
  24. L16:3047 weakly suppresses

Feeds later layers

  1. L19:2440 suppresses
  2. L22:592 suppresses
  3. L20:2781 suppresses
  4. L21:793 suppresses
  5. L18:1655 excites 'and' joining two clauses
  6. L19:2870 suppresses
  7. L21:739 excites 'both X and Y' pairing words
  8. L26:790 suppresses
  9. L25:468 suppresses
  10. L18:1811 excites 'and' linking two descriptive claims
  11. L20:2306 weakly suppresses
  12. L21:1064 weakly suppresses
  13. L20:852 weakly excites
  14. L24:205 weakly excites
  15. L18:2882 weakly suppresses
  16. L20:501 weakly suppresses
  17. L20:68 weakly suppresses
  18. L20:2098 weakly suppresses
  19. L24:478 weakly excites word fragment inside a longer word
  20. L25:2178 weakly suppresses
  21. L18:2041 weakly suppresses
  22. L23:2748 weakly suppresses
  23. L18:149 weakly suppresses
  24. L19:1325 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.

2 values between -3.88 and -3.501 values between -3.50 and -3.136 values between -3.13 and -2.754 values between -2.75 and -2.385 values between -2.38 and -2.006 values between -2.00 and -1.639 values between -1.63 and -1.2525 values between -1.25 and -0.8867 values between -0.88 and -0.50641 values between -0.50 and -0.132495 values between -0.13 and 0.25596 values between 0.25 and 0.6223 values between 0.62 and 1.0010 values between 1.00 and 1.373 values between 1.37 and 1.746 values between 1.74 and 2.123 values between 2.12 and 2.493 values between 2.49 and 2.871 values between 3.24 and 3.621 values between 4.37 and 4.741 values between 10.74 and 11.11offnegativepositiveoften

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. >equals
  2. .defineProperty
  3. 妈
  4. mum
  5. ishing
  6. 昭
  7. /ext
  8. fore
  9. 并
  10. 并向

Suppresses

  1. because
  2. because
  3. 因为
  4. porque
  5. 因为在
  6. 因为他
  7. Because
  8. 因为她
  9. 因为它
  10. 因為