Neuron Explorer

Qwen3-0.6B

commas inside lists of items

A neuron in layer 2 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 responds to commas that separate items in a list, such as 'red, green, and blue' or 'lineups, show-ups, and photographic lineups'. Turning it up made the text wander into unrelated topics, so the list habit did not show up clearly in the output.

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

    , change the -f to -v and add -es⏎In compound words, make the main word plural.<|endoftext|>Mexico has three classifications of holidays: statutory,

  2. #2 peak

    : At the completion of this unit the⏎Chapter Fifteen Summary⏎The police frequently use three procedures to help by witnesses identify criminals: lineups, show-up, and photographic lineups. Lineups involve showing the witness a group of possible suspects

  3. #3 peak

    green and blue flash. Not a myth even in a land of runes, the colorful but elusive glints are caused by atmospheric refraction enhanced by long, low, sight lines and strong atmospheric temperature gradients.⏎Authors & editors

  4. #4 peak

    strikes. And they can dazzle the world with their cute googly eyes.⏎The world looks colorful to our eyes with only three color receptors — red, green, and blue. But for a mantis shrimp, a small marine crustacean

  5. #5 peak

    especially topics that allow students to explore possible majors in communication, business (e.g., marketing, finance), the social sciences (e.g., political science, sociology, global studies), writing, and more. The course will inspire students to think critically

  6. #6 peak

    that is, the tow-plane which pulls them into the air.⏎Within a piston engine, the pistons can be arranged in four ways: radial, in-line, oppositional and "V." The radial engine has pistons arranged in a circle

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 thing is a thing and the thing is a thing and the thing is a thing and the thing is a thing and the thing is a thing and the thing is a thing and the thing is a thing

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

    text in Microsoft Word files.⏎- Heap variables that are freed with free()⏎- Automatic variables left on the stack of languages like C or garbage collected in languages like Java.⏎Byers, Simon. Scalable Exploitation of, and Responses to

  2. #2 peak

    . It provides the means to create engaging user interfaces, implement complex algorithms, and integrate various features seamlessly. By leveraging programming languages such as Java, Swift, or Kotlin, developers can build apps for different platforms like iOS, Android, or cross-platform

  3. #3 peak

    domain names, such as mail services.⏎Why is DNS important?⏎Domain names are fundamental to the Internet. Every Internet-connected device, whether a personal computer, smart phone, or gaming console, looks up each name in the global DNS, and uses

  4. #4 peak

    States with a key source of the high-quality lithium needed to produce lithium-ion batteries used in a range of digital devices and green technologies, from computer tablets and cell phones to electric cars and microgrids.⏎MORE: Salton Sea coverage⏎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. L1:641 excites
  2. L1:698 excites
  3. L1:647 weakly excites
  4. L1:17 weakly suppresses
  5. L1:307 weakly excites
  6. L0:539 weakly suppresses
  7. L1:436 weakly excites
  8. L0:149 weakly excites
  9. L1:1534 weakly suppresses
  10. L1:414 weakly excites programming languages and software names
  11. L1:689 weakly excites
  12. L0:159 weakly suppresses
  13. L1:381 weakly suppresses
  14. L1:2568 weakly excites
  15. L0:301 weakly suppresses
  16. L1:3030 weakly suppresses
  17. L1:853 weakly excites
  18. L0:184 weakly excites
  19. L1:582 weakly suppresses
  20. L1:442 weakly suppresses
  21. L0:1199 weakly excites
  22. L0:638 weakly suppresses
  23. L1:443 weakly suppresses
  24. L1:272 weakly excites

Feeds later layers

  1. L4:100 strongly excites
  2. L6:499 excites
  3. L6:2117 excites
  4. L5:63 excites
  5. L6:123 excites
  6. L3:404 excites
  7. L7:33 suppresses
  8. L7:664 excites
  9. L4:348 suppresses
  10. L6:497 suppresses
  11. L5:477 excites
  12. L3:757 suppresses
  13. L5:42 excites
  14. L5:693 excites
  15. L6:280 excites
  16. L3:666 excites
  17. L5:310 suppresses
  18. L7:120 suppresses
  19. L3:376 excites
  20. L4:508 suppresses
  21. L7:51 excites
  22. L6:111 excites
  23. L9:600 excites
  24. L7:80 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.

2 values between -1.64 and -1.561 values between -1.26 and -1.181 values between -0.88 and -0.803 values between -0.80 and -0.733 values between -0.73 and -0.655 values between -0.65 and -0.5811 values between -0.58 and -0.5012 values between -0.50 and -0.4221 values between -0.42 and -0.3534 values between -0.35 and -0.2787 values between -0.27 and -0.20175 values between -0.20 and -0.12595 values between -0.12 and -0.041803 values between -0.04 and 0.03681 values between 0.03 and 0.11244 values between 0.11 and 0.19110 values between 0.19 and 0.2646 values between 0.26 and 0.3427 values between 0.34 and 0.4112 values between 0.41 and 0.4911 values between 0.49 and 0.578 values between 0.57 and 0.644 values between 0.64 and 0.725 values between 0.72 and 0.792 values between 0.79 and 0.871 values between 0.87 and 0.951 values between 0.95 and 1.021 values between 1.10 and 1.171 values between 1.17 and 1.251 values between 1.33 and 1.40offnegativepositiveoften

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. ')</
  2. anches
  3. repr
  4. HomeController
  5. azeera
  6. .cs
  7. angepicker
  8. 瘙
  9. ữa
  10. imately

Suppresses

  1. 渎
  2. ึ
  3. dign
  4. 現在
  5. aland
  6. lege
  7. INE
  8. mie
  9. ocate
  10. IDA