Neuron Explorer

Qwen3-0.6B

opening parenthesis before an acronym

A neuron in layer 22 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 the open bracket right before a short form or abbreviation, such as '(SpaceX)' or '(SAE)'. When turned up, the model's output became unrelated text with no clear link to acronyms, so the steering evidence does not support the pattern.

Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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

    Space Exploration Technologies (SpaceX)⏎Due to the recent retirement of the space shuttle, NASA⏎now

  2. #2 peak

    thread’s major diameter, minor diameter, pitch diameter, pitch (in case of metric threads) or threads per Inch (TPI in case of Standard American Equivalent (SAE) threads), flank angle, etc. Normally, a GO / NO

  3. #3 peak

    types of external beam radiation, such three-dimensional conformal radiation therapy (3D-CRT) and intensity modulated radiation therapy (IMRT) are

  4. #4 peak

    scanning (SEM) or transmission electron microscopy (TEM) coupled with energy dispersive x-ray spectroscopy (EDX). This technique can provide not only information on the shape of aerosol

  5. #5 peak

    and resource requirements. Developments in GIS have provided agencies with the ability to collaborate more efficiently and effectively. Emergency managers resolve natural disaster challenges using the Comprehensive Emergency Management (CEM) approach. CEM is broken down into four phases: mitigation,

  6. #6 peak

    While computers can already fly a passenger jet better than a trained human pilot, people are still faced with the dangerous yet tedious task of driving automobiles. Intelligent Transportation Systems (ITS) is the field that focuses on integrating information technology with vehicles and transportation infrastructure

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: I have a 3-1-1-3-3-3-3-3-3-3-3-3-3-3-3-3-3-3-3-3-3-3-

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

    74,820 per year for a semi-private room in a nursing home.⏎- $6,965 per month, or $83,580 per year for a private room in a nursing home.⏎- $

  2. #2 peak

    health officials.⏎To date in 2014, thirteen animals have tested positive for rabies in Erie County (five raccoons, five bats, and one each of a cat, skunk and fox).⏎On June 11

  3. #3 peak

    in French in the French part) are available with closed-captioning. Users will also find 164 film clips (80 in English; 84 in French in the French part).⏎These 120 films are

  4. #4 peak

    which means that they were propelled primarily or entirely by sail. The length of the round ships ranged from 26 to 48 feet (8 to 14.7 meters), and they were between 8 and 16 feet

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. L21:1073 excites
  2. L21:2591 suppresses
  3. L20:1845 excites digits inside numbers
  4. L21:569 excites
  5. L21:1359 suppresses
  6. L21:2789 excites
  7. L21:1053 suppresses
  8. L21:1384 excites
  9. L15:187 weakly suppresses
  10. L13:1079 weakly excites
  11. L16:2864 weakly suppresses location or target word before a noun phrase
  12. L21:2069 weakly excites
  13. L20:1906 weakly suppresses
  14. L20:2944 weakly excites
  15. L2:2044 weakly excites
  16. L19:1008 weakly excites
  17. L21:1154 weakly suppresses
  18. L6:1205 weakly excites
  19. L4:119 weakly excites
  20. L4:2364 weakly excites
  21. L8:591 weakly suppresses
  22. L6:1198 weakly suppresses
  23. L0:2287 weakly suppresses
  24. L6:6 weakly suppresses

Feeds later layers

  1. L23:471 weakly suppresses
  2. L24:2602 weakly excites
  3. L23:107 weakly suppresses
  4. L23:54 weakly excites
  5. L23:133 weakly suppresses
  6. L25:2545 weakly excites
  7. L23:1359 weakly excites
  8. L23:1259 weakly excites
  9. L23:2047 weakly suppresses
  10. L23:710 weakly excites
  11. L26:532 weakly excites
  12. L26:834 weakly excites
  13. L25:2208 weakly excites
  14. L23:671 weakly suppresses
  15. L23:231 weakly excites
  16. L23:2394 weakly excites
  17. L25:1718 weakly suppresses
  18. L24:78 weakly suppresses
  19. L27:2696 weakly excites
  20. L23:663 weakly excites
  21. L25:1401 weakly suppresses
  22. L24:2775 weakly excites
  23. L24:2707 weakly suppresses
  24. L25:828 weakly excites

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 -11.22 and -10.491 values between -8.29 and -7.561 values between -6.82 and -6.091 values between -6.09 and -5.364 values between -5.36 and -4.623 values between -4.62 and -3.896 values between -3.89 and -3.1611 values between -3.16 and -2.4324 values between -2.43 and -1.6960 values between -1.69 and -0.96510 values between -0.96 and -0.232915 values between -0.23 and 0.51224 values between 0.51 and 1.2457 values between 1.24 and 1.9726 values between 1.97 and 2.7121 values between 2.71 and 3.4412 values between 3.44 and 4.178 values between 4.17 and 4.902 values between 4.90 and 5.646 values between 5.64 and 6.371 values between 6.37 and 7.105 values between 7.10 and 7.832 values between 7.83 and 8.571 values between 8.57 and 9.302 values between 9.30 and 10.032 values between 12.23 and 12.961 values between 15.90 and 16.631 values between 17.36 and 18.09offnegativepositiveoften

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. lm
  2. 屈
  3. PT
  4. clouds
  5. (single
  6. INCLUDING
  7. 克斯
  8. .ib
  9. (filters
  10. representing

Suppresses

  1. ials
  2. atabase
  3. ernel
  4. ality
  5. 蛄
  6. adow
  7. chosen
  8. ethod
  9. занят
  10. charge