Neuron Explorer

Qwen3-0.6B

blank slots to fill in

A neuron in layer 7 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 blank lines and gaps that someone is meant to fill in, like '___' in a worksheet or '_____' in a sentence. Some of its top hits are not blanks at all, so the pattern is only partly clear. Turning it up or down gave writing that did not reflect the blanks.

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

    -Davis. Demography, July 2, 2014. DOI: 10.1007/s13524-014-0308-7<|endoftext|>Designs and simulation results

  2. #2 peak

    . Its structure is similar to that of the ___2___ and national governments, with an elected mayor as chief of the ___3___ and an elected council that ___4___ the various neighborhoods forming the ___5___ . The ___6___ appoint

  3. #3 peak

    My students ran into several technical challenges; (1) creating a robust shell with adequate wall thickness, (2) how to integrated the speaker grille into the plaster form, and (3) creating a suitable mold for the plaster that could could be

  4. #4 peak

    budget. The council passes city ordinances, sets the tax rate on property, and apportions money among the various___9___. As cities have grown, ____10___ have usually come to represent more than a single neighborhood.⏎3.

  5. #5 peak

    in too little bonding performance of the slurry, and the slurry will not be able to adhere tightly to the foamed plastic net. After demolding, it will have no strength and will easily disassemble. Hang up even if

  6. #6 peak

    No! Your flexible brain is so good at making sense of patterns that it can see two images in one image.⏎The image at left is both a __________________ and __________________.⏎Your brain also solves mysteries in language. Imagine you find the cryptic note

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: "If there is a pattern, and you compare me to, "and if that pattern is true, then there you go, I think the top line of the article, you say, "and if you would say, "take a

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

    *) memchr (str, 'p', strlen(str));⏎printf ("'p' found at position %d.\n", pch-str+1);⏎printf ("'p' not found

  2. #2 peak

    will present new data concerning a stereo photography studio that Dalmau opened in 1854.⏎Download data is not yet available.⏎How to Cite⏎Cuenca, C. (2019). The The “Animated

  3. #3 peak

    fanlike diagram, groups represented by rays that _______________ to the edge of the fan evolved more recently.⏎When organims are classified

  4. #4 peak

    successfully interbreed belong to the same ___________.⏎In the scientific name of the white oak, Quercus alba, Quercus is the __________________.⏎Two groups of organisms that are farther from each other on a cladogram share

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. L6:1039 suppresses
  2. L6:766 excites
  3. L3:1613 weakly suppresses
  4. L6:478 weakly excites
  5. L3:533 weakly excites
  6. L6:540 weakly excites
  7. L6:8 weakly suppresses
  8. L3:2656 weakly excites
  9. L2:1363 weakly suppresses
  10. L4:788 weakly suppresses
  11. L0:1915 weakly excites
  12. L6:2874 weakly excites
  13. L4:2359 weakly excites
  14. L6:13 weakly excites
  15. L3:408 weakly suppresses
  16. L3:42 weakly excites
  17. L5:91 weakly excites
  18. L6:780 weakly excites
  19. L3:372 weakly suppresses
  20. L5:2529 weakly suppresses
  21. L1:1494 weakly excites
  22. L4:316 weakly suppresses
  23. L0:1464 weakly suppresses
  24. L6:1235 weakly excites

Feeds later layers

  1. L9:1449 excites
  2. L8:1228 weakly suppresses
  3. L14:2840 weakly excites
  4. L24:462 weakly suppresses
  5. L24:55 weakly suppresses
  6. L8:1682 weakly suppresses
  7. L14:766 weakly suppresses
  8. L11:1440 weakly excites
  9. L18:1683 weakly suppresses
  10. L17:2514 weakly suppresses
  11. L8:100 weakly excites
  12. L27:1801 weakly excites
  13. L13:693 weakly excites
  14. L13:2855 weakly excites
  15. L8:2598 weakly excites
  16. L22:1947 weakly suppresses
  17. L18:802 weakly excites line break after list items or paragraphs
  18. L22:1384 weakly suppresses
  19. L14:1983 weakly excites
  20. L9:2687 weakly suppresses
  21. L23:1665 weakly suppresses
  22. L17:2278 weakly suppresses
  23. L21:2023 weakly excites
  24. L8:1676 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 -0.66 and -0.631 values between -0.56 and -0.532 values between -0.53 and -0.503 values between -0.50 and -0.473 values between -0.47 and -0.448 values between -0.44 and -0.418 values between -0.41 and -0.3819 values between -0.38 and -0.3423 values between -0.34 and -0.3133 values between -0.31 and -0.2852 values between -0.28 and -0.2583 values between -0.25 and -0.2294 values between -0.22 and -0.19142 values between -0.19 and -0.15195 values between -0.15 and -0.12250 values between -0.12 and -0.09285 values between -0.09 and -0.06365 values between -0.06 and -0.03417 values between -0.03 and 0.01366 values between 0.01 and 0.04344 values between 0.04 and 0.07313 values between 0.07 and 0.10239 values between 0.10 and 0.13187 values between 0.13 and 0.16126 values between 0.16 and 0.19110 values between 0.19 and 0.2382 values between 0.23 and 0.2649 values between 0.26 and 0.2941 values between 0.29 and 0.3223 values between 0.32 and 0.3519 values between 0.35 and 0.387 values between 0.38 and 0.429 values between 0.42 and 0.453 values between 0.45 and 0.482 values between 0.48 and 0.511 values between 0.51 and 0.543 values between 0.57 and 0.61offnegativepositiveoften

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. arehouse
  2. -choice
  3. heit
  4. 平台
  5. 平台上
  6. Onc
  7. ⏎⏎
  8. 决
  9. #⏎⏎
  10. "]),⏎

Suppresses

  1. iscrim
  2. ände
  3. rist
  4. setInput
  5. änd
  6. istically
  7. setAddress
  8. IRST
  9. ルド
  10. heraus