Neuron Explorer

Qwen3-0.6B

the name fragment 'Bon' (Bonn, Bonaparte)

A neuron in layer 23 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 letters 'Bon' at the start of place and person names, most often 'Bonn' in Germany and 'Bonaparte'. It is not a general concept neuron, and turning it up or down did not give clear, related 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

    said during the United Nations Climate Change Conference in Bonn, Germany.⏎Speaking at the conference, China’s top climate negotiator X

  2. #2 peak

    s that came from Saint-Chamond (Loire). He left school of fine arts and architecture of Lyon with honours gained thanks to a silver medal in Bonnefond painting course. After a some year stay in Paris where he became

  3. #3 peak

    the University of Bonn in Germany and colleagues. The team created a three-dimensional interference pattern by shining

  4. #4 peak

    en, as well as an inset at bottom left showing two islands discovered by Cook and Marion du Fresne. One of the few maps in the volume drawn by Bonne's son. Engraved by Andre. Condition: A nice impression with

  5. #5 peak

    traditionally had close economic and cultural relations with both these neighboring regions. Agreeing with the Bon account, they explain that this master, once in Zhang-zhung, combined many

  6. #6 peak

    Haeger's novel Courtesan about Diane de Poitiers and Henri II. From 22 June to 7 July 1815, Bonapartists considered Napoleon

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: if we look at the total cost of the product in the market, we could potentially instead of having a total cost being an indicator, instead using the total cost that is going into the product's production for the product’s market being produced,

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

    Antalya-Türkiye’de doğdu. Lisans öğrenimini MSÜ Deniz Harp Okulunda tamamlamıştır

  2. #2 peak

    ⏎$3 million grant from the Bill & Melinda Gates Foundation will fund development⏎by Eric Klopfer⏎MIT Education

  3. #3 peak

    all remained in place to commemorate the British leaders who championed Jewish rights to a homeland.⏎According to Robert Fulford, a journalist on the Toronto newspaper the Globe & Mail, within a short walk in Jerusalem there are streets called Benjamin Disraeli

  4. #4 peak

    7 to 1947. It is connected to the museum at its top and bottom. © John Grade⏎The new sculpture at the Museum of History and Industry in Seattle

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. L21:1798 suppresses
  2. L20:2799 excites
  3. L20:1750 weakly excites
  4. L0:1139 weakly suppresses
  5. L15:1709 weakly excites
  6. L10:426 weakly excites
  7. L17:1686 weakly excites
  8. L11:2349 weakly excites
  9. L16:2035 weakly excites 'of' after a noun phrase
  10. L0:1767 weakly excites
  11. L20:669 weakly suppresses
  12. L8:14 weakly excites
  13. L12:2602 weakly suppresses
  14. L5:2820 weakly excites
  15. L22:151 weakly suppresses
  16. L21:977 weakly excites
  17. L19:1944 weakly excites
  18. L14:2463 weakly excites
  19. L10:2479 weakly excites
  20. L5:75 weakly excites
  21. L18:1475 weakly suppresses
  22. L2:874 weakly excites
  23. L17:3038 weakly excites
  24. L22:1221 weakly excites

Feeds later layers

  1. L26:2748 weakly excites
  2. L25:1505 weakly excites
  3. L24:478 weakly excites word fragment inside a longer word
  4. L24:967 weakly excites
  5. L24:1660 weakly excites
  6. L27:1119 weakly excites
  7. L25:674 weakly excites
  8. L27:219 weakly excites
  9. L27:1437 weakly excites
  10. L24:298 weakly excites 'would' in hypothetical or past statements
  11. L27:401 weakly excites
  12. L24:2504 weakly excites
  13. L27:693 weakly excites
  14. L26:1516 weakly suppresses
  15. L24:697 weakly suppresses
  16. L25:2359 weakly excites
  17. L26:3004 weakly excites
  18. L24:2802 weakly suppresses
  19. L26:952 weakly suppresses
  20. L24:1678 weakly suppresses
  21. L25:117 weakly excites
  22. L25:1172 weakly suppresses
  23. L24:2974 weakly suppresses
  24. L24:1084 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 -17.81 and -17.221 values between -16.63 and -16.041 values between -14.85 and -14.261 values between -8.93 and -8.341 values between -7.16 and -6.572 values between -6.57 and -5.971 values between -5.97 and -5.382 values between -5.38 and -4.792 values between -4.79 and -4.203 values between -4.20 and -3.617 values between -3.61 and -3.015 values between -2.42 and -1.837 values between -1.83 and -1.2483 values between -1.24 and -0.651515 values between -0.65 and -0.052170 values between -0.05 and 0.5489 values between 0.54 and 1.137 values between 1.13 and 1.722 values between 1.72 and 2.321 values between 2.32 and 2.911 values between 2.91 and 3.501 values between 3.50 and 4.092 values between 4.09 and 4.682 values between 4.68 and 5.281 values between 5.28 and 5.87offnegativepositiveoften

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. return
  2. ized
  3. (return
  4. >Error
  5. nde
  6. attach
  7. close
  8. return
  9. ated
  10. handleError

Suppresses

  1. lector
  2. ajaran
  3. стве
  4. aidu
  5. razier
  6. ERENCE
  7. ollo
  8. цию
  9. acock
  10. cabo