Neuron Explorer

Qwen3-0.6B

preposition introducing a definition or date

A neuron in layer 15 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 words like 'of', 'from' or 'by' that introduce a description of a thing, such as 'a disorder of' or 'characterized by'. It also fires on digits inside birth and death dates in brackets. Turning it up gave replies that drifted into abstract questions without a clear topic.

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

    , and Samm for also getting all or⏎parts of the solution correct.<|endoftext|>See what questions⏎a doctor would ask.⏎Apraxia: A disorder of skilled movement not due to tremors, weakness, akinesia or abnormal tone

  2. #2 peak

    significantly different than previously thought.⏎The results are to be published in the journal Science.⏎The lower mantle comprises 55 percent of the planet by volume and extends from 670 and 2900 kilometers in depth, as defined

  3. #3 peak

    are similar to, or related to, the symptom Apraxia:⏎Read more about causes and Apraxia deaths.⏎Apraxia is a neurological disorder characterized by loss of the ability to execute or carry out learned (familiar) movements,

  4. #4 peak

    for your results and behavior, who or what can guide you? I found wisdom when reading about Confucius and his philosophy.⏎Confucius (551–479BC) was a great teacher and had his teachings written down

  5. #5 peak

    sense of “out of sight, out of mind”. From Wikipedia: The Great Pacific garbage patch, also described as the Pacific trash vortex, is a gyre of marine debris particles in the central North Pacific Ocean located roughly between 135

  6. #6 peak

    power panels can be installed on rooftops or on the ground.⏎- Sunpower solar panels<|endoftext|>By Mortimer J. Adler⏎Aristotle (384 - 322 B.C.) taught common sense to Alexander the nice and

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 number of possible combinations of 2-3-2-2... in order without any mattering in order of repetitions: 1,2 in 3 in 4 in 3 in 2 in ... in order in order

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

    files made available via the university web site. The CMS makes it possible for almost anyone to easily manage web content, without needing to know how to write HTML (hyper-text mark-up language, the typical method for creating web pages). By using a

  2. #2 peak

    to get the⏎"gist" of the message.⏎The Internet Acronym Server⏎This site can be used to look up the meaning of an acronym (abbreviations⏎constructed from the first letters of several words, such as ADA for⏎American

  3. #3 peak

    ⏎Primary Peritoneal Cancer Mays Cancer Center⏎(5 days ago) Primary peritoneal cancer cells start growing in the peritoneum (thin tissue layer that lines the inside of the abdomen). The word “primary” in

  4. #4 peak

    . Which line will be activated when an output device require attention from CPU?<|endoftext|>According to the Oxford Advanced Learners Dictionary, ‘world’ means the earth with all its countries, people and natural features around us. Now, the

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. L14:1050 strongly excites
  2. L13:1885 strongly suppresses
  3. L13:1363 suppresses
  4. L13:2220 suppresses
  5. L8:2518 suppresses
  6. L12:2539 suppresses
  7. L12:1138 excites
  8. L13:116 excites
  9. L14:2587 suppresses
  10. L7:271 excites
  11. L12:1711 suppresses
  12. L13:2058 excites
  13. L14:1019 excites
  14. L11:2184 excites 'way' before a how-to phrase
  15. L13:2904 suppresses
  16. L14:820 excites
  17. L13:1227 excites
  18. L14:1740 weakly excites
  19. L5:569 weakly excites
  20. L12:2892 weakly suppresses
  21. L14:2250 weakly excites
  22. L10:1731 weakly excites
  23. L12:913 weakly excites
  24. L14:2951 weakly excites

Feeds later layers

  1. L19:2391 suppresses
  2. L17:2249 suppresses
  3. L17:2149 suppresses
  4. L17:564 suppresses
  5. L19:929 suppresses
  6. L18:2664 suppresses
  7. L21:826 weakly suppresses
  8. L16:1310 weakly excites
  9. L17:1717 weakly suppresses
  10. L17:876 weakly suppresses
  11. L16:2126 weakly suppresses
  12. L16:1172 weakly suppresses
  13. L16:2192 weakly suppresses
  14. L16:2538 weakly excites
  15. L16:522 weakly suppresses
  16. L19:1663 weakly excites
  17. L16:2151 weakly suppresses
  18. L19:1734 weakly suppresses
  19. L17:2568 weakly suppresses
  20. L19:446 weakly suppresses
  21. L17:1329 weakly suppresses
  22. L16:2224 weakly excites
  23. L22:1781 weakly excites
  24. L18:2562 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 -4.16 and -3.971 values between -2.27 and -2.083 values between -1.90 and -1.713 values between -1.71 and -1.521 values between -1.52 and -1.334 values between -1.33 and -1.148 values between -1.14 and -0.956 values between -0.95 and -0.7619 values between -0.76 and -0.5740 values between -0.57 and -0.39180 values between -0.39 and -0.201423 values between -0.20 and -0.011888 values between -0.01 and 0.18194 values between 0.18 and 0.3776 values between 0.37 and 0.5625 values between 0.56 and 0.7513 values between 0.75 and 0.9414 values between 0.94 and 1.122 values between 1.12 and 1.311 values between 1.31 and 1.502 values between 1.50 and 1.692 values between 1.88 and 2.071 values between 2.63 and 2.821 values between 3.20 and 3.39offnegativepositiveoften

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. -gradient
  2. CellStyle
  3. eld
  4. _SD
  5. 主题教育
  6. xdf
  7. 下的
  8. -cancel
  9. pliers
  10. ipt

Suppresses

  1. 定义
  2. elope
  3. Definition
  4. MISSION
  5. тро
  6. zenia
  7. ńskiego
  8. furnished
  9. 缥
  10. aka