Neuron Explorer

Qwen3-0.6B

the 'conc' word fragment

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 start of words like 'conciliators', 'concise' and 'concave', especially the 'conc' piece. The pushed-up and pushed-down outputs did not show a clear link to this pattern.

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

    are vital right now. Join us!⏎- Animal lovers: Big supermarket brands are directly contributing to this species’ extinction by destroying forests. You can join the #Boycott4Wildlife by sharing information from this website and boycotting brands

  2. #2 peak

    have prompted many people to settle their legal disputes outside the courtroom through a process known as alternative dispute resolution (ADR). Mediators, also known as arbitrators and conciliators, guide the ADR process and help resolve conflicts between disputing parties

  3. #3 peak

    “Right now, cell and medical imaging are at a point comparable to the space<|endoftext|>SAFETY & ESCAPE PLAN⏎BEFORE A VIOLENT INCIDENT OCCURS⏎• Teach your children to call 911.⏎•

  4. #4 peak

    the Catholic Church⏎The Council of Trent (Latin : Concilium Tridentinum), held between 1545 and 1

  5. #5 peak

    the water authority.⏎The discovery of a single fossil creature in a geological stratum of a wrong period, the detection of a single anatomical or physiological fact irreconcilable with origin by descent with modification, would have been destructive of the theory

  6. #6 peak

    in order to earn a complete score, your discussion entries will need to meet the following five main criteria. They must be: written in full sentences, clearly and concisely, answer the questions and/or follow the instructions offered in the prompts,

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: just go just the way way way just the way way just the way way just the way way just the way way just the way way just the way way just the way way way go go go go go go go go go go go go

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

    -cross the land and on Big Moor several stone posts or stoops were erected to guide travellers and traders across the difficult terrain with its bogs and dips, rivulets and hardy moorland vegetation. Mick J

  2. #2 peak

    a scoring system for the musculoskeletal stress marker (MSM) located at the attachment site of the nuchal ligament located on the external occipital protuberance. The principle behind the method is that differences in osteon rem

  3. #3 peak

    were known, such as Hernan Cortes or Diaz del Castillo chronicles. So this study shows disagreements and contradictions of winners.⏎The historian thinks Jose Vasconcelos was wrong when he said that

  4. #4 peak

    Who has power to do everything in this world?⏎Close MATLAB with keyboard⏎Find the treasures in MATLAB Central and discover how the community can help you!⏎Choose a web site to get translated content where available and see local events and offers.

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. L20:748 excites
  2. L22:795 excites
  3. L22:1317 excites unclear: sentence and clause breaks in essays
  4. L17:581 suppresses
  5. L16:704 excites
  6. L16:1461 suppresses
  7. L15:2567 suppresses
  8. L20:304 excites
  9. L8:459 suppresses
  10. L18:1141 weakly excites unclear: mid-sentence words before a noun
  11. L11:309 weakly suppresses
  12. L18:232 weakly suppresses
  13. L3:2811 weakly suppresses
  14. L8:370 weakly excites
  15. L19:2687 weakly suppresses
  16. L17:1963 weakly excites
  17. L8:2754 weakly excites
  18. L21:1001 weakly excites
  19. L3:696 weakly excites
  20. L23:2537 weakly excites
  21. L16:2959 weakly excites
  22. L9:51 weakly excites
  23. L19:1085 weakly excites
  24. L23:99 weakly suppresses

Feeds later layers

  1. L27:1777 weakly excites
  2. L26:2928 weakly suppresses
  3. L27:162 weakly suppresses
  4. L26:1456 weakly excites
  5. L26:1990 weakly suppresses
  6. L25:1311 weakly suppresses
  7. L25:462 weakly suppresses
  8. L25:2833 weakly excites
  9. L26:2494 weakly suppresses
  10. L26:2557 weakly suppresses
  11. L27:1050 weakly suppresses
  12. L27:887 weakly suppresses
  13. L27:1096 weakly excites
  14. L25:2904 weakly suppresses
  15. L26:1555 weakly excites
  16. L26:579 weakly excites
  17. L26:1006 weakly suppresses
  18. L25:2462 weakly suppresses
  19. L25:316 weakly suppresses
  20. L25:2316 weakly excites
  21. L27:720 weakly excites
  22. L27:465 weakly suppresses
  23. L26:1259 weakly excites
  24. L27:2010 weakly 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 -10.16 and -9.641 values between -9.13 and -8.621 values between -8.62 and -8.101 values between -7.59 and -7.082 values between -7.08 and -6.561 values between -6.56 and -6.053 values between -6.05 and -5.543 values between -5.54 and -5.024 values between -5.02 and -4.513 values between -4.51 and -4.008 values between -4.00 and -3.4816 values between -3.48 and -2.9723 values between -2.97 and -2.4640 values between -2.46 and -1.9463 values between -1.94 and -1.43152 values between -1.43 and -0.92421 values between -0.92 and -0.401906 values between -0.40 and 0.11862 values between 0.11 and 0.62202 values between 0.62 and 1.1462 values between 1.14 and 1.6538 values between 1.65 and 2.1626 values between 2.16 and 2.6820 values between 2.68 and 3.1914 values between 3.19 and 3.708 values between 3.70 and 4.226 values between 4.22 and 4.732 values between 4.73 and 5.245 values between 5.24 and 5.762 values between 5.76 and 6.271 values between 6.27 and 6.784 values between 6.78 and 7.303 values between 7.30 and 7.811 values between 8.32 and 8.841 values between 9.35 and 9.862 values between 9.86 and 10.38offnegativepositiveoften

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. ulate
  2. commend
  3. osition
  4. urrence
  5. Th
  6. _initializer
  7. دد
  8. repid
  9. antine
  10. pen

Suppresses

  1. 越多
  2. gotten
  3. go
  4. 之战
  5. goes
  6. iß
  7. 之作
  8. →Get
  9. getList
  10. getVersion