Neuron Explorer

Qwen3-0.6B

inside Italian names and places

A neuron in layer 24 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on the middle pieces of Italian names and places, such as 'Pal-azzo Pitti', 'Pier-antonio' and 'Politecnico di Milano'. It goes strongly negative on Italian words in glossary-style lists. Turning it up or down did not make the model's writing noticeably Italian.

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

    families. The Medici collection began in the 16th century, first in Palazzo Vecchio, then at the Uffizi and finally in Palazzo Pitti with the foundation of the Accademia del Cimento in 1

  2. #2 peak

    1416 and initially worked as a Greek scribe. His study under the famous schoolmaster Vittorino da Feltre gave him a mastery of Latin rare among Greek emigres,

  3. #3 peak

    cent. produced many glorious calligraphic works. Among the best known of these masters were Matteo Contugi, Gianrinaldo Mennio, and Pierantonio S

  4. #4 peak

    of Kenya (January, 2009), a German canton of Switzerland (March, 2009), and the Piemonte-Toscana region of northern/central Italy (images only, February-March,

  5. #5 peak

    2⏎Copyright (c) 2009 Silvia TAVERNINI, Daniele NIZZOLI, Giampaolo ROSSETTI, Pierluigi VIAROLI⏎This work is licensed under a Creative Commons Attribution

  6. #6 peak

    Can self-assembly based technologies offer advantages beyond conventional top-down lithography approaches?⏎A joint effort of the Aalto University of Helsinki, the Politecnico di Milano, and VTT Technical Research Centre of Finland has now demonstrated that it is

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: because of the e-file because of the e- due because of the e- due because of the e-file because because because because because because because because because because because because because because because because because because because because because because because because because because

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

    the frequency.⏎Deutsch: Fabrik / Español: Fábrica / Português: Fábrica / Français: Usine / Italiano: Fabbrica⏎A factory (previously manufactory) or manufacturing plant

  2. #2 peak

    by an adjustable valve and a scaven<|endoftext|>Deutsch: Faktor / Español: Factor - Factor / Français: Facteur - Facteur / Italiano: Fattori - Fattori⏎A factor, a Latin word meaning may

  3. #3 peak

    Is.⏎|Prepared by:||Gordon Rudloff⏎Ohio Department of Agriculture<|endoftext|>Logic & ProofsOpen Learning Initiative⏎Cosa impari in questo corso?⏎Logic & Pro

  4. #4 peak

    Europe, Snam and TIGF) based in six EU Member States plus two renewable gas producing organisations (European Biogas Association and Consorzio Italiano Biogas). The group shares the vision that renewable and low carbon gas,

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. L21:711 strongly excites
  2. L23:352 strongly suppresses
  3. L23:268 suppresses
  4. L22:759 excites
  5. L23:166 suppresses
  6. L23:322 excites
  7. L22:461 excites
  8. L22:700 excites
  9. L21:266 excites
  10. L16:1805 excites
  11. L13:771 weakly suppresses
  12. L22:915 weakly suppresses digits inside historical years
  13. L5:110 weakly suppresses
  14. L6:7 weakly suppresses
  15. L17:640 weakly excites
  16. L8:3 weakly excites
  17. L22:870 weakly suppresses
  18. L1:1431 weakly suppresses
  19. L0:2672 weakly excites
  20. L20:565 weakly excites
  21. L4:2587 weakly excites
  22. L21:437 weakly suppresses
  23. L6:13 weakly excites
  24. L6:8 weakly suppresses

Feeds later layers

  1. L25:744 strongly suppresses
  2. L26:2099 strongly suppresses
  3. L27:2706 strongly suppresses
  4. L26:3021 suppresses
  5. L26:2570 suppresses
  6. L26:2402 suppresses
  7. L26:2462 suppresses
  8. L27:920 suppresses
  9. L26:2428 suppresses
  10. L27:2694 suppresses
  11. L27:2789 suppresses
  12. L27:1824 suppresses
  13. L27:2514 suppresses
  14. L26:1970 suppresses
  15. L27:2690 suppresses
  16. L26:1673 suppresses
  17. L25:2841 suppresses
  18. L26:637 suppresses
  19. L27:2018 suppresses
  20. L27:1699 suppresses
  21. L26:2479 suppresses
  22. L26:1495 suppresses
  23. L26:1126 suppresses the digit 1 in powers of ten
  24. L25:1849 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 -0.52 and -0.451 values between -0.45 and -0.3868 values between -0.04 and 0.03528 values between 0.03 and 0.10576 values between 0.10 and 0.17483 values between 0.17 and 0.24339 values between 0.24 and 0.31323 values between 0.31 and 0.37251 values between 0.37 and 0.44225 values between 0.44 and 0.51164 values between 0.51 and 0.58161 values between 0.58 and 0.65123 values between 0.65 and 0.7292 values between 0.72 and 0.79140 values between 0.79 and 0.8677 values between 0.86 and 0.9252 values between 0.92 and 0.9945 values between 0.99 and 1.0647 values between 1.06 and 1.1336 values between 1.13 and 1.2027 values between 1.20 and 1.2724 values between 1.27 and 1.3426 values between 1.34 and 1.4121 values between 1.41 and 1.4716 values between 1.47 and 1.5417 values between 1.54 and 1.6111 values between 1.61 and 1.686 values between 1.68 and 1.7510 values between 1.75 and 1.824 values between 1.82 and 1.894 values between 1.89 and 1.963 values between 1.96 and 2.023 values between 2.02 and 2.092 values between 2.09 and 2.162 values between 2.16 and 2.23offnegativepositiveoften

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. ữa
  2. ",__
  3. such
  4. icular
  5. 向
  6. Acts
  7. getNode
  8. ĩnh
  9. clubs
  10. 向下

Suppresses

  1. come
  2. ris
  3. proprietà
  4. seg
  5. scop
  6. 对自己的
  7. rice
  8. sia
  9. ecc
  10. pubb