Neuron Explorer

Qwen3-0.6B

opening parenthesis before an acronym

A neuron in layer 23 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 lights up on the opening bracket in text like "Electronic Resources Information Center (ERIC)", where a full name is followed by its short form. Its negative side fires on the same kind of bracket, so the neuron mostly tracks that spot. Turning it up made the output drift into unrelated, rambling text.

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

    L) — can help predict when athletes should return from sports-related concussions, a multicenter, prospective diagnostic study was launched and led by the Advanced Research Core (ARC) of the Concussion Assessment, Research, and Education (CARE)

  2. #2 peak

    o’s Dissertation as a PDF […]⏎The World As It Could Be Curriculum and Resource Guide for Teachers & Organization Leaders is now available through the Education Resources Information Center (ERIC) at http://www.eric.ed.gov/ERICWebPortal/detail?

  3. #3 peak

    funding,” Turner said.<|endoftext|>This article touches on cultural, linguistic, and technical issues that arise in video recording stories of Deaf individuals who use American Sign Language (ASL). Whether you are Deaf or hearing, documenting ASL interviews presents

  4. #4 peak

    have their influence on one another visually mapped. This methodology was named Interpretive Structural Modelling (ISM). Over the next 20 years, the use of ISM to

  5. #5 peak

    made in laboratories around the world. And they are among the goals of more than 100 horizon-stretching research projects housed in the Advanced Measurement Laboratory (AML) Complex at the National Institute of Standards and Technology (NIST).⏎Completed

  6. #6 peak

    patriarch towards the defence of the Holy Land. The struggle ended in the deposition of Dagobert in 1102.⏎<|endoftext|>Agricultural Research Service (ARS) scientists have found environmentally and economically sound uses for

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: how can I be able to understand the future of the world? I think I can understand the future of the world well to some degree, but I feel a little bit of tension. I don't think I can fully understand it, and

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

    kill viruses. Their experiments⏎showed that hypericin was effective in killing many kinds of lentiviruses,⏎especially the equine infectious anemia virus (EIAV), a retrovirus genetically⏎related to the human AIDS virus

  2. #2 peak

    -year mission to study odd clouds high above Earth that shine brightest at night.⏎The U.S. space agency's Aeronomy of Ice in the Mesosphere (AIM) spacecraft rode an Orbital Sciences-built Pegasus XL booster into to orbit

  3. #3 peak

    Engineers (ASHRAE)⏎Structural Engineers Association of Kansas & Missouri (SEAKM)⏎Structural Engineering Institute (SEI)⏎Illuminating Engineering Society (IES)⏎American Society of Plumbing Engineers (ASPE)⏎Society of Fire Protection

  4. #4 peak

    hockey arena so the ice is maintained at just the right temperature.⏎Possible Career Paths and Specializations (See below for respective professional societies)⏎Architectural Engineering Institute (AEI

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. L22:83 excites
  2. L18:3 suppresses
  3. L21:3022 suppresses
  4. L21:1904 suppresses
  5. L22:1545 weakly suppresses
  6. L22:178 weakly suppresses
  7. L22:2768 weakly suppresses
  8. L16:2532 weakly suppresses
  9. L9:2857 weakly excites
  10. L22:1319 weakly excites
  11. L14:1436 weakly suppresses 'and' linking a role to a name
  12. L13:556 weakly excites
  13. L17:2335 weakly excites
  14. L5:486 weakly excites
  15. L5:2567 weakly excites
  16. L20:2035 weakly excites
  17. L21:604 weakly excites
  18. L11:1950 weakly suppresses
  19. L6:2471 weakly excites
  20. L21:10 weakly excites
  21. L20:81 weakly suppresses
  22. L22:978 weakly excites
  23. L12:1781 weakly suppresses
  24. L1:2591 weakly excites

Feeds later layers

  1. L25:446 suppresses
  2. L27:617 excites
  3. L27:266 suppresses
  4. L27:211 excites
  5. L27:1693 excites
  6. L26:483 excites
  7. L27:279 weakly excites
  8. L27:253 weakly excites
  9. L27:243 weakly excites
  10. L24:638 weakly excites
  11. L27:331 weakly excites
  12. L27:268 weakly suppresses
  13. L27:7 weakly excites
  14. L26:2326 weakly excites
  15. L24:2677 weakly suppresses
  16. L27:190 weakly excites
  17. L25:522 weakly excites
  18. L27:1043 weakly excites
  19. L26:249 weakly excites
  20. L26:257 weakly excites
  21. L24:1001 weakly excites
  22. L24:2701 weakly excites
  23. L24:3032 weakly suppresses
  24. L24:1856 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 -18.05 and -16.611 values between -10.86 and -9.421 values between -7.98 and -6.544 values between -6.54 and -5.117 values between -5.11 and -3.676 values between -3.67 and -2.2396 values between -2.23 and -0.793708 values between -0.79 and 0.6564 values between 0.65 and 2.087 values between 2.08 and 3.528 values between 3.52 and 4.962 values between 4.96 and 6.401 values between 7.84 and 9.271 values between 23.65 and 25.091 values between 38.03 and 39.47offnegativepositiveoften

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. esc
  2. emachine
  3. erate
  4. oration
  5. AT
  6. emer
  7. еп
  8. ackers
  9. ASC
  10. AB

Suppresses

  1. pcs
  2. ppt
  3. FS
  4. pcs
  5. Kai
  6. 濉
  7. PCS
  8. pj
  9. uai
  10. CCS