Neuron Explorer

Qwen3-0.6B

word before a verb of action or effect

A neuron in layer 14 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 word just before an action verb in a plain explanatory sentence, such as "have also [produced]", "can [cause]" or "to [produce]". It is weaker on the word "of" after nouns like "copy of" or "member of", where it turns negative. Turning it up did not give a clear theme.

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

    travel and movement of goods, the importance of maintaining robust biosecurity measures at a country’s borders cannot be overstated. These measures are designed to prevent the entry of potentially devastating pests, diseases and invasive species, which could have serious economic, environmental

  2. #2 peak

    -grey on the face and chest. They have gorgeous blue flight feathers and a lighter green tinge on the underside of their tails. Captive breeding programs have also produced a variety of beautiful color mutations in Quakers.⏎There is a blue hybrid Qu

  3. #3 peak

    is now an important part of a city's defense. We can no longer depend on historical weather patterns to be a guide. Sea-level rise and climate change are creating new and dangerous weather systems that can be deadly.⏎These developments have brought new energy

  4. #4 peak

    Encourage them to explore scientific concepts through fun experiments, build structures using household materials, or engage in fun projects. There are numerous online resources and educational platforms that offer interactive STEM activities suitable for everyone from kindergarten to high school such as Khan Academy,

  5. #5 peak

    new types of DNA molecules called DNA-based polymers. The DNA-based polymers, often referred to as designer genes, are used in a variety of ways to produce new proteins, genes

  6. #6 peak

    the balance of oxygen flow and heat loss needs to be carefully balanced, spontaneous ignition is not common but is still a risk that needs to be managed Some businesses have had unsafe work practices for many years before they have a fire caused by spontaneous ignition.⏎

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 "the" of the new, the of the, the of, and of the, and of the, when "the" of the, and of the, and of the, and of the, and of the, and of

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

    4. […]⏎Jacques le Moyne de Morgues (c. 1533–1588) was a French artist and member of Jean Ribault’s expedition to the New World. His depictions of Native American

  2. #2 peak

    difficult. The study addressed the issue by modifying the ALA<|endoftext|>Note: I published a previous version of this article on The Change Blog (now called Possibility of Change) in

  3. #3 peak

    later version published by the Free Software Foundation; with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the license is included in the page “GNU Free Documentation License”.⏎The copyright

  4. #4 peak

    Qualitative Research: State of the Art and the Future Glossary References Author index Subject index About the Author.⏎(source: Nielsen Book Data)⏎'The fourth edition of Uwe Flick's "Introduction to Qualitative Research" remains the most comprehensive

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. L13:203 excites
  2. L10:272 suppresses
  3. L13:1161 excites
  4. L11:153 suppresses
  5. L12:434 suppresses
  6. L13:1427 excites
  7. L6:388 excites
  8. L7:23 suppresses
  9. L2:2539 excites
  10. L9:2198 suppresses
  11. L8:216 weakly excites
  12. L8:45 weakly suppresses
  13. L4:76 weakly suppresses
  14. L12:1393 weakly excites
  15. L5:146 weakly excites
  16. L8:328 weakly suppresses
  17. L11:1168 weakly suppresses
  18. L6:914 weakly suppresses
  19. L12:372 weakly suppresses
  20. L12:336 weakly suppresses
  21. L13:44 weakly excites inside web addresses and citations
  22. L2:832 weakly suppresses
  23. L2:740 weakly suppresses
  24. L9:74 weakly suppresses

Feeds later layers

  1. L16:896 excites
  2. L16:1320 suppresses
  3. L20:95 excites
  4. L16:2063 excites
  5. L19:278 suppresses
  6. L18:156 excites
  7. L19:157 suppresses
  8. L16:269 suppresses
  9. L17:410 excites
  10. L16:786 excites
  11. L18:606 excites
  12. L20:738 suppresses
  13. L19:152 suppresses
  14. L18:63 excites
  15. L19:204 suppresses
  16. L19:363 suppresses unclear: table pipes and split word fragments
  17. L20:60 suppresses
  18. L16:297 excites
  19. L16:340 excites
  20. L19:1129 excites
  21. L25:16 suppresses
  22. L20:161 excites
  23. L16:176 suppresses
  24. L17:57 excites

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 -3.86 and -3.711 values between -3.71 and -3.551 values between -3.55 and -3.402 values between -3.40 and -3.251 values between -2.33 and -2.184 values between -2.18 and -2.033 values between -2.03 and -1.875 values between -1.87 and -1.727 values between -1.72 and -1.575 values between -1.57 and -1.4210 values between -1.42 and -1.268 values between -1.26 and -1.1122 values between -1.11 and -0.9622 values between -0.96 and -0.8131 values between -0.81 and -0.6538 values between -0.65 and -0.5073 values between -0.50 and -0.35142 values between -0.35 and -0.19526 values between -0.19 and -0.041639 values between -0.04 and 0.11775 values between 0.11 and 0.26371 values between 0.26 and 0.42118 values between 0.42 and 0.5734 values between 0.57 and 0.7218 values between 0.72 and 0.8810 values between 0.88 and 1.0311 values between 1.03 and 1.189 values between 1.18 and 1.337 values between 1.33 and 1.493 values between 1.49 and 1.642 values between 1.64 and 1.792 values between 1.79 and 1.955 values between 1.95 and 2.102 values between 2.10 and 2.25offnegativepositiveoften

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. plements
  2. cripts
  3. plies
  4. ell
  5. مل
  6. criptions
  7. ellite
  8. >{"
  9. atel
  10. cribed

Suppresses

  1. 的味道
  2. Fabric
  3. овое
  4. 放出
  5. 控股
  6. 被抓
  7. atrav
  8. roma
  9. 这家
  10. the