Neuron Explorer

Qwen3-0.6B

unclear: scattered mid-phrase words

A neuron in layer 14 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 a grab bag of words that sit in the middle of a phrase, with no clear shared theme. When it was turned up, the model's text wandered into unrelated topics, so there is no clear story about what it does.

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

    growing love handles despite diet and exercise. Also deteriorating athletic performance, a general sense of fatigue or falling asleep right after dinner can all be signs of low testosterone, but they are often written

  2. #2 peak

    as being a List of Users.⏎Sometimes, it’s also handy to define a list of multiple Entities and Structures.⏎Until 184.108.40.2

  3. #3 peak

    68k-palmos-coff-gcc -v should report its version number to be (at time of writing) 22.214.171.124-kgpd-07109

  4. #4 peak

    and would like to contribute something to the GNU project, please consider volunteering for one of these projects. If you are seriously contemplating work, please write to `firstname.lastname@example.org' to coordinate with other

  5. #5 peak

    unkel⏎starring Jim Witter,⏎featuring Ian Tanner⏎“Time it was, and what a time it was”...⏎These opening lyrics to Simon

  6. #6 peak

    1931, John and Jack, two of the caretakers of the Imaginarium Geographica, discover a plea for help on a medieval manuscript - which seems to have been written by their friend, Hugo Dyson! When

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: the person who is not in the right, the right, and the right is not in the right, and the right is not, and the right is not, and the right is not, and the right is not, and the right

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

    even after advice⏎- Be obsessed with body image and shape. They often fail to recognize the ill-effects of excess weight loss.⏎The most telling signs about anorexia can be picked around the dining table as they try

  2. #2 peak

    .⏎Edwin Lacey was a relative of the Revolutionary War general and the son of Isaac and Ruth (nee Twining). Lacey was a tall, gaunt, red-headed bachelor, a Quaker, and a farmer. He was an

  3. #3 peak

    baby warm.⏎An old custom of washing a newborn soon after birth is being discounted with the knowledge that infants were becoming sick. Instead, a baby should be gently dried with a towel and placed next to the mother's skin. Washing should wait until

  4. #4 peak

    for a description). So you'll have to call this macro 5 times to get the 5 characters.⏎Remember that when you print a string it should be null

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. L5:1812 weakly excites
  2. L9:1766 weakly excites
  3. L2:444 weakly excites
  4. L7:152 weakly excites
  5. L9:2509 weakly suppresses
  6. L13:1124 weakly excites
  7. L5:2501 weakly suppresses
  8. L11:2290 weakly excites
  9. L12:2254 weakly excites
  10. L13:1549 weakly suppresses
  11. L4:273 weakly suppresses
  12. L10:2144 weakly excites
  13. L12:3013 weakly excites
  14. L4:1066 weakly excites
  15. L5:2700 weakly suppresses
  16. L8:2501 weakly suppresses 'iv' inside herbivores, energy fuel words
  17. L9:932 weakly excites
  18. L10:1837 weakly excites
  19. L1:2672 weakly suppresses
  20. L6:799 weakly excites
  21. L3:2791 weakly suppresses
  22. L11:2306 weakly excites
  23. L2:1950 weakly excites
  24. L12:2208 weakly excites

Feeds later layers

  1. L18:1509 weakly excites
  2. L22:2685 weakly excites
  3. L22:2116 weakly excites
  4. L21:2047 weakly excites
  5. L21:2013 weakly excites
  6. L18:2617 weakly excites
  7. L27:1770 weakly excites
  8. L20:2829 weakly excites
  9. L25:303 weakly excites
  10. L21:1545 weakly excites
  11. L22:2344 weakly excites number after 'account for' or 'over'
  12. L22:2247 weakly excites
  13. L21:2946 weakly excites
  14. L20:307 weakly suppresses
  15. L22:1625 weakly excites
  16. L23:563 weakly excites
  17. L18:1530 weakly suppresses
  18. L16:900 weakly excites
  19. L24:2891 weakly excites
  20. L20:1299 weakly excites
  21. L22:2740 weakly suppresses
  22. L19:1902 weakly excites
  23. L15:2423 weakly excites
  24. L19:790 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 -2.04 and -1.961 values between -1.88 and -1.801 values between -1.57 and -1.492 values between -1.49 and -1.411 values between -1.26 and -1.182 values between -1.18 and -1.101 values between -1.10 and -1.022 values between -1.02 and -0.952 values between -0.95 and -0.873 values between -0.87 and -0.793 values between -0.79 and -0.715 values between -0.71 and -0.644 values between -0.64 and -0.569 values between -0.56 and -0.4811 values between -0.48 and -0.4022 values between -0.40 and -0.3266 values between -0.32 and -0.25115 values between -0.25 and -0.17343 values between -0.17 and -0.09962 values between -0.09 and -0.011582 values between -0.01 and 0.07375 values between 0.07 and 0.14170 values between 0.14 and 0.2293 values between 0.22 and 0.3051 values between 0.30 and 0.3825 values between 0.38 and 0.4619 values between 0.46 and 0.5311 values between 0.53 and 0.6110 values between 0.61 and 0.692 values between 0.69 and 0.778 values between 0.77 and 0.853 values between 0.85 and 0.921 values between 0.92 and 1.002 values between 1.00 and 1.08offnegativepositiveoften

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. 掌门
  2. Switch
  3. umer
  4. clusion
  5. recess
  6. 帝王
  7. universities
  8. millennium
  9. discoveries
  10. aro

Suppresses

  1. uated
  2. iffin
  3. ava
  4. rated
  5. Pvt
  6. 醪
  7. .Views
  8. veled
  9. laughter
  10. haft