Neuron Explorer

Qwen3-0.6B

'of' after superlatives and 'the X of'

A neuron in layer 9 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 word 'of' when it comes after a superlative or a noun that is being ranked, such as 'the largest nature reserve of' or 'the biggest holiday of the year'. Turning it up did not produce a clear theme in the written text.

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

    Cite This Page:<|endoftext|>National Park Hohe Tauern⏎The National Park Hohe Tauern (in German spelled Nationalpark) is the largest nature reserve of the Alps and the second-largest national park of Europe - it is 1.

  2. #2 peak

    suggests that, according to the Greeting Card Association, one billion cards are now sent each year, making Valentine’s Day ‘the second-largest card-sending holiday of the year, surpassed only by Christmas.’⏎Having previously heard hearsay that the

  3. #3 peak

    REMINDER : The Renaissance was a cultural and artistic movement of the 15th and 16th centuries originating in Italy, between

  4. #4 peak

    and perhaps 10 to 20 whites).⏎What’s known as the 1921 Tulsa Race Massacre is generally considered the single worst incident of racial violence in U

  5. #5 peak

    election, hacktivist group Anonymous Africa attacked and closed down 50 websites, including those associated with the ruling Zanu PF party.⏎Most Famous Group of Hacktivism:⏎Two of the most widely known groups are Anonymous and Lul

  6. #6 peak

    physician immediately:⏎- Heavy vaginal bleeding⏎- Passing blood clots⏎- Moderate or severe cramping⏎- Unusual vaginal discharge⏎- Chills or fever of 100.4 degrees Fahrenheit or higher⏎The physical and emotional pain

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 the data of the number of people are using the website, and I know that the website is using the Google Ads and the other ad campaigns are set. I need to know if I should check the ad campaigns' website for the

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

    - morning sickness⏎More About Morning Sickness⏎Morning sickness often begins 4 to 6 weeks after conception. It may continue until the 4th month of pregnancy. Some women have morning sickness during their entire pregnancy. This

  2. #2 peak

    2020. If statistics are to be believed, it can be concluded that cloud computing is the future of the IT Domain.⏎CASE STUDY: Federation of Indian Chambers of Commerce and Industry (FICCI)⏎F<|endoftext|>A joint study

  3. #3 peak

    prejudice which continue to affect people of Afrikan descent today."⏎On 25 March 2007, bicentenary of the Abolition of the Slave Trade, then Prime Minister Tony Blair of Britain publicly stated that slavery was

  4. #4 peak

    , in the northern beaches area of Sydney (as well as the Western suburbs of Sydney), and the South Coast of NSW. The LDAA is in the process

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. L3:315 excites 'of' after a noun phrase
  2. L4:4 suppresses
  3. L6:357 suppresses
  4. L5:363 excites
  5. L2:895 weakly suppresses
  6. L5:1828 weakly excites
  7. L4:534 weakly suppresses
  8. L8:2334 weakly excites
  9. L2:2833 weakly excites
  10. L4:1686 weakly excites
  11. L6:7 weakly suppresses
  12. L3:1379 weakly excites
  13. L0:2726 weakly excites
  14. L6:1627 weakly suppresses
  15. L4:1098 weakly suppresses
  16. L4:1675 weakly suppresses
  17. L1:725 weakly suppresses "to" in "referred to"
  18. L1:1050 weakly suppresses
  19. L4:2648 weakly excites
  20. L6:13 weakly excites
  21. L1:951 weakly excites
  22. L4:648 weakly suppresses
  23. L0:1332 weakly suppresses
  24. L6:2931 weakly suppresses

Feeds later layers

  1. L19:715 excites
  2. L16:1088 excites 'of' after a noun phrase, as in 'of course'
  3. L16:2035 excites 'of' after a noun phrase
  4. L17:317 excites 'of' after a noun phrase
  5. L11:2207 excites "of" after a noun in trademark and title phrases
  6. L14:2417 weakly excites
  7. L18:1546 weakly excites
  8. L15:368 weakly excites
  9. L18:1997 weakly suppresses
  10. L24:148 weakly suppresses
  11. L10:469 weakly excites 'of' after a noun phrase
  12. L16:1350 weakly suppresses
  13. L25:934 weakly excites
  14. L22:353 weakly suppresses
  15. L17:1835 weakly excites
  16. L18:282 weakly suppresses
  17. L10:1712 weakly excites
  18. L11:805 weakly excites
  19. L20:1972 weakly excites 'of' in 'of course'
  20. L17:222 weakly excites
  21. L19:1119 weakly excites
  22. L26:217 weakly excites
  23. L10:346 weakly excites "each" in "each other" phrases
  24. L19:1437 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 -2.70 and -2.491 values between -1.86 and -1.641 values between -1.64 and -1.431 values between -1.43 and -1.224 values between -1.22 and -1.013 values between -0.79 and -0.5855 values between -0.58 and -0.37506 values between -0.37 and -0.162279 values between -0.16 and 0.05959 values between 0.05 and 0.2768 values between 0.27 and 0.488 values between 0.48 and 0.694 values between 0.69 and 0.904 values between 0.90 and 1.124 values between 1.12 and 1.331 values between 1.33 and 1.541 values between 1.54 and 1.751 values between 1.75 and 1.961 values between 1.96 and 2.181 values between 2.18 and 2.391 values between 3.24 and 3.451 values between 3.87 and 4.081 values between 4.30 and 4.511 values between 5.15 and 5.361 values between 5.57 and 5.78offnegativepositiveoften

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. mer
  2. gener
  3. medium
  4. minor
  5. course
  6. s
  7. **
  8. 조
  9. กา
  10. icon

Suppresses

  1. 因为她
  2. 对她
  3. heets
  4. irez
  5. 告诉她
  6. 下来的
  7. chaft
  8. neau
  9. reffen
  10. 于是我