Neuron Explorer

Qwen3-0.6B

unclear: pieces of split words

A neuron in layer 1 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 words that have been chopped into fragments, most strongly in 'laissez-faire' and in surnames like 'Dupriez'. Turning it up made the model's output drift into loose, unrelated sentences, so it does not clearly steer any topic.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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

    Pine Ridge, he advocated for peace with the European Americans⏎Date de dernière mise à jour : 05/07/2021⏎Ajouter un commentaire<|endoftext|>By DANIEL SUDDEATH⏎NEW ALBANY —

  2. #2 peak

    of placing the retaliatory use of force under objective control.⏎When I say “capitalism,” I mean a full, pure, uncontrolled, unregulated laissez-faire capitalism — with a separation of state and economics, in the same way

  3. #3 peak

    .”12⏎However, banning research is also unlikely to be desirable as rogue research could still continue.2 This may lead to a worse outcome than a laissez-faire approach because clandestine science has the capacity to drive cold wars for decades

  4. #4 peak

    (at least some of them).⏎Eventually, Golway argues, Tammany Hall introduced the notion that the government had a social welfare role to play. Laissez-faire capitalism needed some help, and the corrupt politicians of that era helped push

  5. #5 peak

    3(3), pages 315-320.⏎- Angus Deaton & Olivier Dupriez, 2011.⏎"Purchasing Power Parity Exchange Rates for

  6. #6 peak

    .⏎(2) Internal data – Laboratoires URGO – QIPI VLU 2019 / 4⏎(3) Améliorer la qualité du système de santé et maîtriser les dépenses : propositions

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 been thinking about something. I have been thinking about something. I have been thinking about something. I have been thinking about something. I have been thinking about something. I have been thinking about something. I have been thinking about something

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

    BPA. Glass bottles, however, also leach chemicals, including aluminum, explains a study by Clemens Reimann of the Norges Geologiske Undersøkelse in Norway and colleagues. Aluminum is sometimes added to soda-l

  2. #2 peak

    3(3), pages 315-320.⏎- Angus Deaton & Olivier Dupriez, 2011.⏎"Purchasing Power Parity Exchange Rates

  3. #3 peak

    Col. Stand Watie, with one regiment of Cherokees, to move into the neutral land and

  4. #4 peak

    de Vereenigde Doopsgezinde Gemeente to Amsterdam, 2 vols. Amsterdam: Uitgegeven en ten geschenke aangeboden door den Kerkeraad dier Gemeente, 1

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. L0:1301 suppresses
  2. L0:178 suppresses
  3. L0:1224 suppresses
  4. L0:1193 weakly excites
  5. L0:1 weakly suppresses
  6. L0:894 weakly excites
  7. L0:489 weakly excites
  8. L0:1921 weakly excites
  9. L0:1167 weakly suppresses
  10. L0:220 weakly suppresses
  11. L0:75 weakly suppresses
  12. L0:429 weakly excites
  13. L0:2641 weakly suppresses
  14. L0:84 weakly suppresses
  15. L0:1468 weakly suppresses
  16. L0:30 weakly suppresses
  17. L0:1836 weakly suppresses
  18. L0:2094 weakly excites
  19. L0:25 weakly excites
  20. L0:485 weakly excites
  21. L0:2542 weakly excites
  22. L0:48 weakly excites
  23. L0:91 weakly suppresses
  24. L0:386 weakly suppresses

Feeds later layers

  1. L2:765 excites
  2. L2:93 excites
  3. L3:104 excites
  4. L3:110 suppresses
  5. L27:149 weakly suppresses
  6. L6:819 weakly excites
  7. L13:612 weakly excites
  8. L3:970 weakly excites
  9. L3:2807 weakly excites
  10. L25:307 weakly excites
  11. L27:108 weakly suppresses
  12. L4:266 weakly suppresses
  13. L23:1780 weakly excites
  14. L19:1788 weakly excites
  15. L6:31 weakly excites
  16. L9:413 weakly suppresses
  17. L3:74 weakly excites
  18. L9:584 weakly suppresses
  19. L4:420 weakly excites
  20. L4:528 weakly excites
  21. L7:2490 weakly excites
  22. L5:928 weakly excites
  23. L2:26 weakly excites
  24. L23:1561 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 -0.43 and -0.411 values between -0.39 and -0.371 values between -0.35 and -0.333 values between -0.33 and -0.311 values between -0.31 and -0.293 values between -0.29 and -0.274 values between -0.27 and -0.253 values between -0.25 and -0.2310 values between -0.23 and -0.225 values between -0.22 and -0.206 values between -0.20 and -0.1812 values between -0.18 and -0.1619 values between -0.16 and -0.1420 values between -0.14 and -0.1227 values between -0.12 and -0.1050 values between -0.10 and -0.0877 values between -0.08 and -0.06121 values between -0.06 and -0.04217 values between -0.04 and -0.02653 values between -0.02 and -0.001159 values between -0.00 and 0.02697 values between 0.02 and 0.04346 values between 0.04 and 0.05178 values between 0.05 and 0.0798 values between 0.07 and 0.0959 values between 0.09 and 0.1137 values between 0.11 and 0.1323 values between 0.13 and 0.1515 values between 0.15 and 0.1717 values between 0.17 and 0.1912 values between 0.19 and 0.2112 values between 0.21 and 0.234 values between 0.23 and 0.258 values between 0.25 and 0.273 values between 0.27 and 0.282 values between 0.28 and 0.302 values between 0.30 and 0.322 values between 0.32 and 0.34offnegativepositiveoften

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. =DB
  2. десь
  3. babel
  4. =db
  5. acular
  6. =Value
  7. =""><
  8. Weber
  9. 的一款
  10. parable

Suppresses

  1. yk
  2. ULT
  3. dom
  4. 人在
  5. DEAL
  6. hari
  7. istically
  8. ettle
  9. alties
  10. -domain