Neuron Explorer

Qwen3-0.6B

hyphen inside compound words

A neuron in layer 20 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 hyphen in words joined by a dash, like "above-mentioned," "at-your-own-pace" and "low-and middle-income." It is mostly silent elsewhere. Turning it up did not produce any obvious theme in the text that followed.

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

    COVID-19 diagnostic based on micro/nano technology and how to tackle the above-mentioned challenges.⏎The exploration of surfaces, interfaces, nanostructures and thin films using

  2. #2 peak

    honor the format of exploring hands-on activities, learning about artwork in our galleries, and making an inspired art project.⏎As always, Kid Quest is an at-your-own-pace art adventure. In this case, take all the time you like

  3. #3 peak

    The currently known individual allergens of the above-mentioned allergy triggers from the arthropod realm are summarized and their potential use in

  4. #4 peak

    engineers will hold their breath as the newest Landsat satellite is launched into space aboard an Atlas V rocket. Landsat 8 carries one of the most sophisticated Earth-observing systems, but it also carries the weight of a

  5. #5 peak

    Not at all, says Wolf. Instead, she argues, “The great challenge now is to learn how to use both print and digital mediums to their best<|endoftext|>- A Tale of Two Brothers: Haiti’s Other Revolutions⏎Sévère Cour

  6. #6 peak

    of the online version is:⏎Several examples of recently-constructed buildings are shown, all of them designed with the latest 3-D modeling

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: theseseseseseseseseseseseseseseseseseses.This is the first thing I have done, andI have the first callout callout callout

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

    Proceedings of the Iowa Academy of Science, 31(1), 131-132.⏎Available at: https://scholarworks.uni.edu/pias/vol31/iss1/25<|endoftext|>New images

  2. #2 peak

    .⏎Avesoro CRUSHING AND PROCESSING OPERATIONS AT NEW⏎2 Nov 2015 Ore crushing operations recommenced on 28 October 2015, following the “

  3. #3 peak

    Build a broad network for yourself. Seek to enlarge you network of associates, with the goal of expanding your interests and opportunities.⏎COMMUNICATE. NEGOTIATE. APPRECIATE.<|endoftext|>West Nile Virus⏎News of an epidemic

  4. #4 peak

    call came to late. Fedorets fired at McConnell with his MiG-15’s three autocannon (a single Nudelman N-37 37 mm gun and two Nudelman-Rikhter NR-2

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. L2:404 strongly excites
  2. L19:89 strongly excites
  3. L1:217 strongly suppresses
  4. L3:508 strongly suppresses the final zero of decades like '1980s'
  5. L19:43 strongly excites
  6. L16:3 strongly suppresses
  7. L5:97 strongly suppresses
  8. L5:621 excites
  9. L16:7 suppresses
  10. L16:0 suppresses
  11. L3:395 suppresses
  12. L4:490 suppresses decade digit in years like 1930s
  13. L16:1 excites
  14. L15:123 suppresses
  15. L0:132 suppresses
  16. L8:3004 excites
  17. L0:2215 suppresses
  18. L2:801 suppresses
  19. L6:715 excites
  20. L2:761 suppresses
  21. L2:422 excites
  22. L18:853 suppresses
  23. L0:579 suppresses
  24. L4:384 suppresses

Feeds later layers

  1. L22:265 strongly excites
  2. L24:248 strongly excites
  3. L25:422 strongly excites
  4. L24:318 strongly suppresses
  5. L25:231 strongly suppresses
  6. L26:1008 strongly suppresses
  7. L27:2408 strongly suppresses
  8. L26:2003 strongly excites hyphen joining two word parts
  9. L26:254 strongly suppresses
  10. L27:185 strongly excites
  11. L27:1463 strongly excites
  12. L22:2384 strongly excites
  13. L26:1776 strongly excites
  14. L27:2439 strongly excites slash inside web addresses
  15. L27:145 strongly excites dots inside web addresses
  16. L27:1019 strongly suppresses
  17. L26:725 excites
  18. L27:170 excites unclear: spaces in mixed-language text
  19. L27:981 excites
  20. L27:2657 excites
  21. L23:474 excites
  22. L23:393 suppresses
  23. L25:36 excites
  24. L25:535 excites commas inside number lists

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 -10.81 and -9.961 values between -4.03 and -3.191 values between -3.19 and -2.347 values between -2.34 and -1.50474 values between -1.50 and -0.653082 values between -0.65 and 0.2022 values between 0.20 and 1.0435 values between 1.04 and 1.8936 values between 1.89 and 2.7452 values between 2.74 and 3.5851 values between 3.58 and 4.4336 values between 4.43 and 5.2840 values between 5.28 and 6.1230 values between 6.12 and 6.9713 values between 6.97 and 7.815 values between 7.81 and 8.667 values between 8.66 and 9.516 values between 9.51 and 10.351 values between 10.35 and 11.201 values between 12.05 and 12.892 values between 12.89 and 13.741 values between 13.74 and 14.581 values between 17.12 and 17.971 values between 18.82 and 19.661 values between 19.66 and 20.511 values between 22.20 and 23.05offnegativepositiveoften

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. RIPTION
  2. /he
  3. цин
  4. ẩn
  5. ائن
  6. ForResult
  7. Є
  8. Offsets
  9. 见过
  10. арь

Suppresses

  1. s
  2. y
  3. i
  4. t
  5. n
  6. a
  7. m
  8. S
  9. d
  10. sing