Neuron Explorer

Qwen3-0.6B

capital N starting a name or word

A neuron in layer 27 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 reacts to a capital N that begins a new piece of a word, most often the first letter of a surname or a place name. When it was turned up, the model's writing did not become clearly related to names; it drifted into odd filler words like 'expected' and 'overcast'.

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

    not land a steady job. A handful of boys from up and down Senegal’s coast. The captain, Mamour Ndiaye, who always goes to sea wearing the same tank top in the pattern

  2. #2 peak

    Employment Opportunity for Women) Act (AA Act) was passed, with the Affirmative Action Agency established to<|endoftext|>During a conversation with ChileCarne, José Naranjo, Veterinary Doctor and international consultant in Animal Health and Public Health, provided

  3. #3 peak

    ades are vulnerable to high temperatures, though it seems that these critters would be able to acclimatize to increasing temperatures in their natural habitat,” says Ricardo Nieves, an author of<|endoftext|>January 10, 2002

  4. #4 peak

    need to take time and look around. It is amazing what the world has to teach us!⏎Yes, I am also fully agree with you both (Rob and Naperlou). In this case we can take the best example from nature i

  5. #5 peak

    iers often took part and Louis' nickname, The Sun King, came from his performance as Apollo, the Greek god of the sun, in the Ballet de la Nuit in

  6. #6 peak

    al & Behavioral Pediatrics.⏎Previous news in sleep:<|endoftext|>In the Dutch city of Oss, 60 miles southeast of Amsterdam, there's a highway named N329. During the day, N329 is a stretch of

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:)))); )))); )))); )))); )))); )))); )))); )))); )))); )))); )))); )))); )))); )))); )))); )))); )))); )))); )))); )))); )))); )))); )))); ]))); ]))); ]))); )))); ]))); ]))); ]))); ]))); )))); ]))); ]))); ]))); ]))); )))); ]))); ]))); ]))); ]))); ]))); ]))); ]))); ]))); ]))); ]))); ])));

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

    in Scotland originated with Anglo-Norman knights bearing the de Beauville or, more commonly, de Boyville name

  2. #2 peak

    origins as a surname are quite uncertain. David de Olifard is the progenitor of the House of Oliphant. He was one of the many Anglo-Norman nobles that were invited northward by the early Norman kings of Scotland.

  3. #3 peak

    is credited with discovering the islands in 1493. They were held by European colonial powers such as Spain, Britain, the Netherlands, France and Denmark-Norway one after the other. The Islands were purchased by the United States from

  4. #4 peak

    to prepare for immediate launch. Silo crews on land were notified. Their targets would be the major cities of North America and western Europe. They knew the American/NATO forces would respond in kind.⏎The

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. L24:708 suppresses
  2. L26:876 suppresses
  3. L22:2299 excites
  4. L26:1564 excites
  5. L26:2121 suppresses
  6. L20:212 excites
  7. L2:1923 suppresses
  8. L17:391 excites
  9. L5:499 weakly excites
  10. L5:2586 weakly excites
  11. L20:1120 weakly excites
  12. L0:26 weakly excites
  13. L22:2314 weakly excites
  14. L23:697 weakly excites
  15. L5:57 weakly suppresses
  16. L23:1532 weakly suppresses
  17. L21:2959 weakly excites
  18. L6:13 weakly excites
  19. L8:10 weakly excites
  20. L18:569 weakly excites
  21. L8:7 weakly excites
  22. L23:2139 weakly excites
  23. L24:2239 weakly excites
  24. L21:475 weakly excites

Feeds later layers

None: this is the last layer.

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.

4 values between -9.70 and -5.3723 values between -5.37 and -1.043875 values between -1.04 and 3.291 values between 3.29 and 7.611 values between 7.61 and 11.941 values between 11.94 and 16.271 values between 29.25 and 33.571 values between 33.57 and 37.901 values between 159.05 and 163.38offnegativepositiveoften

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. heel
  2. ئة
  3. _expected
  4. dpi
  5. acr
  6. closed
  7. Due
  8. vọng
  9. toch
  10. noun

Suppresses

  1. ])));⏎
  2. ])))⏎
  3. ))));⏎
  4. ])))
  5. "))))⏎
  6. ))))⏎
  7. ))))⏎⏎
  8. ))))
  9. ]]];⏎
  10. ']]);⏎