Neuron Explorer

Qwen3-0.6B

'-ment' word endings split mid-word

A neuron in layer 23 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 the middle pieces of longer words that end in 'ment', such as 'assessment', 'allotment' and 'abandonment'. It is quiet, or goes the other way, on the word 'half' in phrases like 'at half past'. Turning it up made the model produce odd words ending in 'ment', so the pattern shows up in the output too.

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

    LIMITED WARRANTY*⏎All spray foam installations are carried out by Lapolla / AirTight licensed installers.⏎*See manufacturer guidelines and requirements.⏎HOME ENERGY ASSESSMENT⏎A home energy assessment helps owners determine where their house is losing energy and

  2. #2 peak

    the acronym Family ASSESSMENT—Structure, Socioeconomic, Sickness, Nutritional, Educational,

  3. #3 peak

    impossible to give your garden or allotment the care it needs to thrive. And even those without the luxury of their

  4. #4 peak

    its former glory. Today, the West Baden Hotel is a National Historic Landmark and continues to attract visitors from all over the world.⏎Decline And Abandonment⏎The West Baden Hotel was once known as the “Eighth Wonder

  5. #5 peak

    writer and translator. Born in London, she now lives in a small village near Oxford. She enjoys growing flowers, fruit and vegetables in her garden, down the allotment and as a volunteer with therapeutic gardening charity Restore. She is especially interested in

  6. #6 peak

    otechnic" reactions, i.e. shock- induced reactions in heterogeneous ma- terial mixtures.⏎- ISOTOPIC SEPARATION AND ENRICHMENT⏎- “ballotechnic” weapons, possibly pure fusion weapons.⏎-

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: how to way in which way we way way way way way way way way way way way way way way way way way way way way way way way way way way way way way way way way way way way way way way way way way

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

    I opened the window, it was raining heavily and people⏎were hurrying along the streets⏎2. Every day the boss enters the office⏎at half past nine⏎3. My friend will ring me up at eight o’clock

  2. #2 peak

    window, it (to rain) heavily and people (to hurry) along the streets. 2. Every day the boss (to enter) the office at half past

  3. #3 peak

    name appears, select Grades. On the Grades page, select the cell that displays the average grade in your course. The Course Activity Related to Grades report appears when half of your students have received more than one grade.⏎If your class enrollment is in

  4. #4 peak

    asked to suggest further objects and can actively participate by uploading photographs of their own objects that have a local or global appeal.⏎History of the World events⏎During February half-term, there is a guided walk which will tell the story of the ten objects

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. L16:120 weakly suppresses
  2. L3:2594 weakly excites
  3. L6:238 weakly suppresses
  4. L19:1851 weakly excites
  5. L20:1616 weakly excites
  6. L2:84 weakly excites
  7. L17:127 weakly suppresses the word 'would' before a verb
  8. L20:2889 weakly suppresses
  9. L2:2347 weakly excites
  10. L1:1273 weakly suppresses
  11. L5:2577 weakly suppresses
  12. L5:19 weakly suppresses
  13. L4:992 weakly suppresses
  14. L6:2745 weakly excites
  15. L18:511 weakly suppresses
  16. L9:2975 weakly excites
  17. L16:3017 weakly excites
  18. L22:1871 weakly suppresses
  19. L20:2371 weakly excites
  20. L17:2620 weakly suppresses
  21. L0:2462 weakly suppresses
  22. L19:246 weakly excites
  23. L16:155 weakly suppresses
  24. L11:2252 weakly excites

Feeds later layers

  1. L24:3045 excites
  2. L25:1061 excites
  3. L26:1533 weakly excites
  4. L27:1809 weakly suppresses
  5. L24:1671 weakly suppresses
  6. L24:979 weakly suppresses
  7. L25:1087 weakly excites word right before a number value
  8. L27:364 weakly suppresses
  9. L27:992 weakly suppresses
  10. L25:1020 weakly excites
  11. L24:2895 weakly suppresses
  12. L26:1682 weakly suppresses
  13. L24:2878 weakly suppresses
  14. L26:2998 weakly suppresses
  15. L27:1543 weakly excites
  16. L26:102 weakly excites
  17. L26:1754 weakly excites opening parenthesis before an acronym
  18. L24:1523 weakly excites
  19. L24:1940 weakly suppresses
  20. L24:1605 weakly excites
  21. L27:2373 weakly suppresses
  22. L27:422 weakly excites
  23. L26:1578 weakly excites
  24. L27:2559 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 -72.88 and -70.631 values between -54.88 and -52.631 values between -25.63 and -23.382 values between -21.13 and -18.881 values between -18.88 and -16.642 values between -16.64 and -14.394 values between -12.14 and -9.895 values between -9.89 and -7.6413 values between -7.64 and -5.3919 values between -5.39 and -3.1475 values between -3.14 and -0.893730 values between -0.89 and 1.3628 values between 1.36 and 3.619 values between 3.61 and 5.867 values between 5.86 and 8.114 values between 8.11 and 10.362 values between 10.36 and 12.611 values between 12.61 and 14.863 values between 14.86 and 17.11offnegativepositiveoften

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. ment
  2. MENT
  3. ments
  4. amento
  5. amiento
  6. amentos
  7. ements
  8. ament
  9. AMENT
  10. ement

Suppresses

  1. way
  2. -way
  3. way
  4. Way
  5. 径
  6. 之一
  7. ONGL
  8. WAY
  9. 徑
  10. WAY