Neuron Explorer

Qwen3-0.6B

'of' after a place or role noun

A neuron in layer 12 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 lights up on the word 'of' when it follows a noun about location or role, like 'site of', 'place of residence', or 'medium of'. When it was turned up, the written text did not clearly change in theme, and it drifted into repeated phrases.

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

    ized times|⏎|Editors||Kerry John KENNEDY, Chi Kin John LEE|⏎|Place of Publication||Abingdon, Oxon|⏎|Publication status||E-pub

  2. #2 peak

    or prevent platelets from coming together. Whenever there is an injury in your body, platelets are sent to the site of the injury, where they clump together to form a blood clot. This stops

  3. #3 peak

    fellow (1994)⏎- IJCAI Award for Research Excellence (1995)⏎- Katherine, Peter, and Barbara⏎- Place of Birth⏎- Milwaukee, Wisconsin⏎- Fields of Expertise⏎- [“

  4. #4 peak

    components of the bloodstream, coordinating with platelet cells to stop bleeding at the site of an injury by forming into a gel-like plug. Platelets are normal cell fragments

  5. #5 peak

    student’s school career” (cited in Crees & Blackledge, 2010:103).⏎The use of English as the medium of class instruction in mathematics classrooms, however, creates another barrier for students, besides mastering

  6. #6 peak

    that this circumstance affords a general indication of the part of the wall upon which they labored, such places being on that side of the city nearest their place of abode; the only apparent exception being, perhaps, where they repaired more than

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 don't have any problems with the right and left, but I feel like it's like a "barrel in the middle" that is on the other side, but there is a problem with the right, left, up, and

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

    who discovered it in 1880. William Sutherland was one of several Scots who explored the country in the 19th century. He ended a wandering life (he had earlier fought under Garibaldi in Italy) when he

  2. #2 peak

    of 5. In California, every year, 75 families lose a child under the age of 5 to drowning. Help protect your child from these preventable tragedies.⏎Drownings

  3. #3 peak

    (by about 5-8 points).⏎Find aerobic exercise you enjoy and can do daily, such as walking or jogging, swimming, cycling, high-intensity interval training, or even dancing. Also include strength training exercise at least two days per

  4. #4 peak

    logo, it’s a good idea to transfer it to copyright. However, when you sell something like software that you could sell again to another client, think twice before parting with your economic rights.<|endoftext|>The USA Basketball Player Development Curriculum has been

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. L10:1211 suppresses
  2. L11:1392 excites 'of' in fixed technical phrases
  3. L8:500 suppresses 'to' after verbs like continue or help
  4. L6:2287 suppresses
  5. L9:2583 excites
  6. L6:685 excites
  7. L9:731 suppresses
  8. L9:282 weakly suppresses
  9. L9:1476 weakly excites
  10. L4:1653 weakly excites
  11. L6:1313 weakly suppresses
  12. L5:766 weakly suppresses
  13. L4:1499 weakly suppresses
  14. L11:292 weakly suppresses
  15. L8:3041 weakly excites
  16. L6:2877 weakly suppresses
  17. L5:1639 weakly excites
  18. L5:1759 weakly excites
  19. L11:429 weakly suppresses
  20. L11:1881 weakly excites
  21. L1:1589 weakly suppresses
  22. L4:2748 weakly suppresses
  23. L1:2363 weakly suppresses
  24. L10:1011 weakly suppresses

Feeds later layers

  1. L17:1225 strongly excites
  2. L18:529 strongly excites
  3. L13:1335 excites 'of' inside fixed noun phrases
  4. L13:1100 excites
  5. L19:2052 excites
  6. L19:1582 excites "of" after an abstract noun
  7. L17:2495 excites
  8. L14:1737 excites
  9. L21:215 suppresses
  10. L17:2192 excites
  11. L15:1344 excites
  12. L19:776 excites
  13. L13:1845 excites
  14. L24:480 excites
  15. L17:871 excites
  16. L18:597 suppresses
  17. L19:1840 excites
  18. L20:1444 excites
  19. L13:1623 excites
  20. L21:656 excites
  21. L21:535 excites
  22. L22:280 suppresses
  23. L19:1396 excites
  24. L17:546 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.

2 values between -3.17 and -2.991 values between -1.77 and -1.593 values between -1.41 and -1.243 values between -1.24 and -1.065 values between -1.06 and -0.897 values between -0.89 and -0.7126 values between -0.71 and -0.54112 values between -0.54 and -0.36499 values between -0.36 and -0.191552 values between -0.19 and -0.011398 values between -0.01 and 0.16175 values between 0.16 and 0.3448 values between 0.34 and 0.5226 values between 0.52 and 0.6914 values between 0.69 and 0.8710 values between 0.87 and 1.046 values between 1.04 and 1.229 values between 1.22 and 1.392 values between 1.39 and 1.572 values between 1.57 and 1.741 values between 1.74 and 1.922 values between 1.92 and 2.091 values between 2.09 and 2.271 values between 2.27 and 2.451 values between 2.45 and 2.621 values between 2.62 and 2.801 values between 3.67 and 3.85offnegativepositiveoften

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. 用餐
  2. 发展战略
  3. aming
  4. การ
  5. 开采
  6. ในการ
  7. 创作
  8. shipping
  9. 仓储
  10. sampling

Suppresses

  1. ibles
  2. вед
  3. 자가
  4. ables
  5. ators
  6. imeters
  7. adoras
  8. -minded
  9. perse
  10. オリジ