Neuron Explorer

Qwen3-0.6B

unclear: verbs before an object

A neuron in layer 15 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on words that come just before an object or complement, such as 'answer', 'from' and 'in' in sentences about school or careers. It turns negative on food and diet words like 'oatmeal' and 'diet soda' in health advice. The pattern is weak and mixed, and turning it up did not give a clear theme.

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

    However, showing you is much cooler than explaining it.⏎Thanks for Subscribing!⏎You'll need to confirm your email, so please check your inbox.⏎(click anywhere to close)<|endoftext|>According to findings of a major HIV prevention trial presented Monday,

  2. #2 peak

    University of Minnesota. To ask her a gardening question, call 612-673-7793 and leave a message. She will answer questions in this column only.⏎© 20

  3. #3 peak

    African American to serve in the Illinois Senate.Biography Roberts was born August 20, 1866 in Decatur, Michigan. He graduated from high school at 17 and became a teacher. He then chose to take

  4. #4 peak

    the office, the position was variously designated as "attorney for the state" or "state's attorney." The attorney general was for many years appointed, first by the legislature and then by the governor. Georgia's Constitution of 18

  5. #5 peak

    approach to eating? Try these tips to get started down the path to health and longevity!⏎- Aim for at least one fully vegetarian meal a week and go from there⏎If you are new to this concept it may be a little intimidating to completely

  6. #6 peak

    Students need to complete two parts of the text with provided words. After that, they have to write 5 comprehension questions concerning the text. They can ask and answer them in pairs. Finally, there is an open discussion on the topic of clown

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: and if I, if I not if, and if, and not if and not if and If, and if and If and the only if and and, and and and if and and and and and and and and and and and and

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

    wasn’t proven entirely. If you are planning to eat oats or oatmeal now, make sure that you purchase whole oats and steel-cut oats, as packaged oatmeal contains flour and

  2. #2 peak

    such as 3 to 4 glucose tablets, a glass of nonfat milk, or 4 ounces (½ cup) of juice or regular soda. Diet soda will not help at all. Chocolate, cookies, and other fatty sweets will not

  3. #3 peak

    iles. These animals are ectothermic, meaning they gather body heat from their surroundings. There are only three species of reptiles in the park, with a fourth suspected to be present.⏎The common garter

  4. #4 peak

    . Protocol numbers internet assigned numbers authority.⏎Protocol cisco commands ios configuration 61 cheat sheets tagged with network. This number is assigned automatically by the os, manually by the. They are associated with tcpip network connections and might be described as

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. L14:2162 suppresses
  2. L13:152 excites
  3. L14:2360 excites
  4. L14:2245 suppresses
  5. L14:1761 excites
  6. L13:706 excites
  7. L14:236 excites 'and' joining two parallel subjects
  8. L14:845 weakly suppresses
  9. L13:2059 weakly suppresses
  10. L13:1172 weakly suppresses
  11. L9:738 weakly excites
  12. L14:2858 weakly excites
  13. L13:2769 weakly excites
  14. L14:1093 weakly excites
  15. L13:2435 weakly suppresses
  16. L14:1777 weakly suppresses
  17. L13:1273 weakly suppresses
  18. L11:2160 weakly excites
  19. L14:696 weakly suppresses
  20. L13:2829 weakly suppresses
  21. L13:1014 weakly suppresses
  22. L5:101 weakly excites
  23. L1:502 weakly excites
  24. L0:1149 weakly excites

Feeds later layers

  1. L16:1838 strongly suppresses
  2. L18:1187 suppresses
  3. L17:1640 suppresses
  4. L18:2774 suppresses
  5. L16:768 excites
  6. L17:449 suppresses
  7. L16:826 excites
  8. L16:1802 suppresses
  9. L16:669 excites
  10. L16:1646 suppresses
  11. L16:1442 suppresses
  12. L16:2802 suppresses
  13. L18:845 suppresses
  14. L21:930 suppresses
  15. L21:1861 suppresses
  16. L17:207 excites
  17. L16:1166 suppresses
  18. L19:1995 suppresses comma after a subject phrase
  19. L17:1826 suppresses
  20. L18:843 suppresses
  21. L20:2592 suppresses
  22. L17:1394 suppresses
  23. L19:2498 suppresses
  24. L23:449 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 -2.09 and -2.002 values between -2.00 and -1.911 values between -1.54 and -1.452 values between -1.45 and -1.362 values between -1.36 and -1.261 values between -1.26 and -1.174 values between -1.17 and -1.081 values between -1.08 and -0.997 values between -0.99 and -0.909 values between -0.90 and -0.808 values between -0.80 and -0.717 values between -0.71 and -0.6214 values between -0.62 and -0.5318 values between -0.53 and -0.4434 values between -0.44 and -0.3456 values between -0.34 and -0.25189 values between -0.25 and -0.16479 values between -0.16 and -0.071523 values between -0.07 and 0.02900 values between 0.02 and 0.12386 values between 0.12 and 0.21159 values between 0.21 and 0.3056 values between 0.30 and 0.3920 values between 0.39 and 0.488 values between 0.48 and 0.586 values between 0.58 and 0.672 values between 0.67 and 0.767 values between 0.76 and 0.851 values between 0.85 and 0.941 values between 0.94 and 1.041 values between 1.04 and 1.131 values between 1.13 and 1.221 values between 1.40 and 1.491 values between 1.49 and 1.59offnegativepositiveoften

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. 羯
  2. YPE
  3. ande
  4. SEARCH
  5. .And
  6. REAK
  7. anter
  8. search
  9. alli
  10. VEL

Suppresses

  1. '][]
  2. 做不到
  3. voir
  4. gợi
  5. cannot
  6. does
  7. misses
  8. vanished
  9. ]()
  10. if