Neuron Explorer

Qwen3-0.6B

'scripting' before 'language', language loops

A neuron in layer 25 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron is hard to pin down. Its strongest negative reactions come on the word 'scripting' in phrases like 'server-side scripting language'. Turning it down made the model get stuck repeating the word 'language', so the neuron seems tied to the word 'language' in programming contexts.

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

    ְדָה (vÿ’ikkavÿda), the Niphal cohortative; coming after the perfect tenses with vav (ו) consecutives expressing the future, this cohort

  2. #2 peak

    to tropical storm status, with maximum sustained winds of 65 mph (105 km/h).⏎On September 14, Humberto had reached post-tropical cyclone status, with winds at 40 mph (65 km

  3. #3 peak

    are very cheap and reliable in cost when we compared with the OP-Amp applications in the same areas. These ICs are used as an astable and monost

  4. #4 peak

    >> Project. Select General >> Empty Project and type in the project name.⏎- Under folder Source, right-click Select Add >> New Item.⏎- Under Code, Select C++ File (.cpp), however, when you type in the name use a

  5. #5 peak

    79. The withdrawal of RBGAN occurred because it was superseded by the newer BGAN technology.⏎BGAN signal acquisition requires line-of-site with the geostationary satellite and requires the user to have a compass and a general idea of

  6. #6 peak

    ⏎- Cancer (malignant neoplasms)⏎- Chronic lower respiratory disease⏎- Accidents (unintentional injuries)⏎- Stroke (cerebrovascular diseases)⏎- Alzheimer’s disease⏎- Influenza and pneumonia

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: language and language language language language language language... language language language language language language language language language language language... language language language language language language language language language language language language... language language language language language language language language language language language language language...

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

    the most widely used server-side scripting language⏎At time of this writing, PHP is the most used server-side language

  2. #2 peak

    , nearly every databases – MySQL, ORACLE, PostgreSQL, Sybase, Informix to name a few.⏎What is PHP?⏎- PHP is a server-side scripting language that means its code is executed on

  3. #3 peak

    although there are other possibilities. ⏎Students are expected to develop their understanding of programming beyond the skill level (that is, beyond the particular commands and syntax of a specific language) and develop an understanding of the general principles and characteristics of programming and programming

  4. #4 peak

    features is essential. This helps developers make informed decisions regarding the choice of programming languages, frameworks, and development tools. Secondly, adhering to coding best practices

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. L23:443 suppresses
  2. L21:652 suppresses
  3. L15:921 excites
  4. L20:1634 suppresses
  5. L21:3028 suppresses
  6. L21:269 excites
  7. L19:552 suppresses
  8. L6:2790 excites
  9. L21:1206 excites
  10. L8:67 weakly suppresses
  11. L7:2300 weakly excites
  12. L23:2267 weakly excites
  13. L24:851 weakly suppresses
  14. L9:49 weakly suppresses
  15. L9:411 weakly excites
  16. L9:2711 weakly excites
  17. L5:135 weakly suppresses
  18. L4:1128 weakly suppresses
  19. L8:2765 weakly excites
  20. L7:2721 weakly excites
  21. L20:429 weakly excites
  22. L7:833 weakly excites
  23. L9:2033 weakly excites
  24. L21:1793 weakly suppresses

Feeds later layers

  1. L27:1855 strongly suppresses
  2. L26:523 weakly suppresses
  3. L27:908 weakly excites
  4. L26:1427 weakly excites
  5. L26:3036 weakly excites
  6. L27:2840 weakly suppresses
  7. L27:2147 weakly excites
  8. L27:2553 weakly suppresses capital T starting a word
  9. L26:1744 weakly suppresses
  10. L27:1216 weakly suppresses
  11. L26:2385 weakly excites
  12. L26:2568 weakly excites
  13. L26:2948 weakly excites
  14. L27:2508 weakly excites
  15. L26:1812 weakly excites
  16. L27:1473 weakly excites
  17. L26:1599 weakly excites
  18. L27:2979 weakly excites
  19. L26:3028 weakly suppresses
  20. L26:1196 weakly excites
  21. L27:1448 weakly suppresses
  22. L27:629 weakly excites
  23. L26:1136 weakly excites
  24. L27:834 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.

3 values between -8.49 and -7.801 values between -6.40 and -5.714 values between -5.71 and -5.013 values between -5.01 and -4.322 values between -4.32 and -3.626 values between -3.62 and -2.936 values between -2.93 and -2.2321 values between -2.23 and -1.53171 values between -1.53 and -0.841726 values between -0.84 and -0.141794 values between -0.14 and 0.55115 values between 0.55 and 1.2516 values between 1.25 and 1.957 values between 1.95 and 2.646 values between 2.64 and 3.346 values between 3.34 and 4.034 values between 4.03 and 4.732 values between 4.73 and 5.432 values between 5.43 and 6.122 values between 6.12 and 6.822 values between 6.82 and 7.512 values between 8.21 and 8.912 values between 9.60 and 10.302 values between 10.30 and 10.991 values between 11.69 and 12.391 values between 15.86 and 16.561 values between 18.65 and 19.34offnegativepositiveoften

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. software
  2. ASTER
  3. 软件
  4. resign
  5. класс
  6. akte
  7. uty
  8. umer
  9. software
  10. {*

Suppresses

  1. 语言
  2. language
  3. Language
  4. languages
  5. language
  6. Language
  7. 的语言
  8. 語言
  9. _language
  10. lang