Neuron Explorer

Qwen3-0.6B

'of' in titles and place names

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 lights up on the word 'of' when it links a job title or place to a name, as in 'Bishop of London' or 'City of Opelika'. Its opposite is the word 'from' after a noun, as in 'residues from'. Turning it up did not make the model's writing clearly about titles or places.

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

    could hardly be found to bury the dead, and many thousand bodies were carelessly thrown into mere pits dug in the open fields.⏎Ralph Stratford, Bishop of London, shocked at these unsanctified interments, in his zeal to

  2. #2 peak

    1472, Niccolò Perotti (1430-1480), secretary to Cardinal Bessarion and archbishop of Siponto, donated to Sassoferrato, the city of his birth

  3. #3 peak

    century.⏎The site lay waste for about 150 years but, in the mid-11th century, Robert, bishop of Coutances, undertook the rebuilding of the cathedral in the Romanesque or

  4. #4 peak

    with the study of literature.<|endoftext|>The two Supreme Court cases, Jones v. City of Opelika 316 U.S. 584 (1

  5. #5 peak

    C. (Image via Zach Montellaro on Flickr, CC BY-NC-SA 2.0)⏎The two Supreme Court cases, Jones v. City of Opelika 316 U.S. 584 (1

  6. #6 peak

    10, 2011⏎At 3pm on the 10th June 1540 Thomas Cromwell, the Earl of Essex and Henry VIII’s right-hand man, was in the council chamber at Westminster

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: the user of the previous year. I am not sure of the previous year. It's like the user of the previous year... the user of the previous year... the user of the previous year... the user of the previous year... 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

    personally how these gardens evolved under their care. The walled garden at Parham was probably created in the 18th century and paths and brick walls survive from that period. It must always have been a productive garden, supplying food and flowers

  2. #2 peak

    of cesium due to leakage from the Fukushima power plant, depending on origin. See: Arsenic, Methylmercury.⏎Shellfish may contain residues from chemicals that have filtered or run off into the water in which they inhabit. Consumers

  3. #3 peak

    of the old Gold Hill Hotel where Nancy Melton became operator⏎One of the tombstones in the Gold Hill cemetery, Ephraim Mauney, was from Stanly County⏎After the death of John Sr., and Joseph Sr. ,

  4. #4 peak

    distance, are the only remaining relics of this green cathedral.⏎More cemeteries, designed by John Bergmans,

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. L4:192 excites
  2. L6:2937 excites 'by' after a title or author credit
  3. L4:2859 weakly excites
  4. L0:1660 weakly suppresses
  5. L2:1147 weakly suppresses
  6. L3:1027 weakly excites
  7. L7:1970 weakly excites
  8. L0:993 weakly excites
  9. L8:451 weakly suppresses
  10. L5:1922 weakly suppresses 'to' in 'X to Y' phrases
  11. L8:16 weakly suppresses
  12. L4:2404 weakly excites
  13. L4:2839 weakly suppresses
  14. L4:2717 weakly excites
  15. L4:1721 weakly suppresses the gap before a decade or century
  16. L6:454 weakly excites
  17. L6:1604 weakly excites 'of' after 'that' in comparisons
  18. L8:685 weakly suppresses
  19. L7:2248 weakly excites
  20. L7:1054 weakly suppresses
  21. L5:2984 weakly suppresses
  22. L8:2397 weakly excites
  23. L6:2664 weakly excites
  24. L5:939 weakly suppresses

Feeds later layers

  1. L14:2007 suppresses
  2. L16:2196 excites
  3. L17:550 excites
  4. L15:2682 weakly suppresses
  5. L13:775 weakly suppresses
  6. L16:2087 weakly suppresses
  7. L12:2889 weakly suppresses
  8. L13:558 weakly suppresses
  9. L22:1326 weakly excites
  10. L13:894 weakly suppresses 'of' after a person's name
  11. L16:2780 weakly excites period after 'Dr' abbreviation
  12. L23:1931 weakly suppresses
  13. L10:1054 weakly suppresses 'from' after movement or transfer verbs
  14. L20:1804 weakly suppresses
  15. L12:2671 weakly suppresses
  16. L17:189 weakly suppresses
  17. L16:2530 weakly suppresses
  18. L14:1094 weakly suppresses
  19. L12:2850 weakly suppresses
  20. L14:2460 weakly suppresses
  21. L19:2298 weakly suppresses
  22. L10:132 weakly suppresses
  23. L18:1979 weakly suppresses
  24. L16:593 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 -5.27 and -5.101 values between -4.74 and -4.572 values between -3.86 and -3.681 values between -3.51 and -3.331 values between -2.98 and -2.801 values between -2.80 and -2.621 values between -2.62 and -2.451 values between -2.45 and -2.271 values between -2.27 and -2.092 values between -1.92 and -1.741 values between -1.74 and -1.577 values between -1.57 and -1.392 values between -1.39 and -1.217 values between -1.21 and -1.0411 values between -1.04 and -0.8611 values between -0.86 and -0.688 values between -0.68 and -0.5166 values between -0.51 and -0.33296 values between -0.33 and -0.151765 values between -0.15 and 0.031396 values between 0.03 and 0.20275 values between 0.20 and 0.3837 values between 0.38 and 0.564 values between 0.56 and 0.733 values between 0.73 and 0.912 values between 0.91 and 1.091 values between 1.26 and 1.441 values between 1.44 and 1.613 values between 1.61 and 1.79offnegativepositiveoften

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. weise
  2. パス
  3. .addCell
  4. ppers
  5. pute
  6. oppers
  7. 信じ
  8. ContentPane
  9. .addChild
  10. /topics

Suppresses

  1. SourceType
  2. aida
  3. haunt
  4. ю
  5. toa
  6. ainting
  7. 来源
  8. 吉利
  9. AEA
  10. clid