Neuron Explorer

Qwen3-0.6B

opening parenthesis after a term

A neuron in layer 18 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 fires on an open bracket placed right after a name or term that is about to be shortened, as in Sustainable Development Goals (SDGs). It is very similar to a neighbouring neuron that does the same job, and turning it up produced unrelated 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

    Brix Orolfo⏎Winner, Geology rocks drawing competition 2011, secondary school category<|endoftext|>The recently adopted United Nation’s Sustainable Development Goals (SDGs) set the new global development agenda for the next 15 years

  2. #2 peak

    sensitivity, says Cowley: "This work really does point to pursuing that as a<|endoftext|>(NaturalNews) Ohio farmers are more reluctant to plant genetically modified organisms (GMOs) in their fields than their counterparts in other Midwestern states,

  3. #3 peak

    now at least 877 times the background extinction rate over geological time.⏎BioScience, published monthly, is the journal of the American Institute of Biological Sciences (AIBS; www.aibs.org). BioScience is a forum for integrating

  4. #4 peak

    , and be creative and get the gift accordingly.⏎- Creativity wins:⏎You<|endoftext|>How integrated learning systems benefit teachers⏎Subscribe to our newsletter⏎Integrated Learning Systems (ILS) are solutions that deliver instructional content. They may be offline or online

  5. #5 peak

    -23/the-future-of-automation-in-the-mining-industry/8550636<|endoftext|>Service Oriented Architecture⏎Service-oriented architecture (SOA) is an evolution of distributed computing based on the

  6. #6 peak

    didn’t realize it at the time, but he was already practicing socially responsible investing.⏎What is Socially Responsible Investing?⏎Socially responsible investing (sometimes also called Ethical Investing) is exactly what it sounds like. It’s

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: I have a 2012, 2012, 2012, 2012, 2012, 2012, 2012, 20

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

    33) has a total length of 125,982 bp, and it is located in the 7q31 locus of chromosome 7. MET is transcribed into a 6,641 bp mature

  2. #2 peak

    were issued.⏎The Revenue Act of 1862⏎On July 1, 1862, President Abraham Lincoln signed the Revenue Act of 1862 into law, to help fund the Civil War.⏎Narc

  3. #3 peak

    's Board of Missions of the Interior, located in Chicago (WBMI), the Woman's Board of Missions for the Pacific, located in San Francisco (WBMP), and the Woman's Board of Missions for the Pacific Islands, located

  4. #4 peak

    intended for use in massive concrete structures, such as large gravity dams, where the temperature rise resulting from heat generated during hardening must be minimized. Type V: Type V portland cement is used in concrete exposed to severe sulfate environments (princip

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. L17:1794 suppresses
  2. L17:2806 suppresses
  3. L15:2704 suppresses
  4. L15:185 suppresses
  5. L17:2002 suppresses
  6. L17:2444 suppresses
  7. L17:440 excites
  8. L17:1191 suppresses
  9. L16:1181 excites
  10. L17:239 suppresses
  11. L15:1901 suppresses
  12. L14:595 excites
  13. L17:532 excites
  14. L17:1022 suppresses
  15. L17:2264 suppresses
  16. L16:2591 suppresses
  17. L17:1948 suppresses
  18. L17:873 excites
  19. L14:707 excites
  20. L17:987 suppresses
  21. L16:1045 suppresses
  22. L13:2908 excites
  23. L15:1684 excites
  24. L12:2132 suppresses

Feeds later layers

  1. L19:1673 excites
  2. L20:867 excites
  3. L20:1189 excites
  4. L22:106 excites
  5. L19:653 suppresses
  6. L19:225 suppresses
  7. L24:104 excites
  8. L19:1428 excites
  9. L20:681 weakly suppresses
  10. L19:762 weakly suppresses
  11. L21:451 weakly suppresses
  12. L19:355 weakly suppresses
  13. L20:1458 weakly excites
  14. L23:133 weakly suppresses
  15. L23:348 weakly excites
  16. L20:294 weakly suppresses
  17. L19:940 weakly excites
  18. L22:73 weakly suppresses
  19. L20:697 weakly excites
  20. L19:1055 weakly excites
  21. L20:1372 weakly excites
  22. L19:731 weakly excites
  23. L19:392 weakly suppresses
  24. L20:872 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.

2 values between -3.50 and -3.242 values between -3.24 and -2.993 values between -2.99 and -2.731 values between -2.73 and -2.488 values between -2.48 and -2.223 values between -2.22 and -1.9712 values between -1.97 and -1.7215 values between -1.72 and -1.4619 values between -1.46 and -1.2133 values between -1.21 and -0.9569 values between -0.95 and -0.70134 values between -0.70 and -0.44326 values between -0.44 and -0.191987 values between -0.19 and 0.07896 values between 0.07 and 0.32240 values between 0.32 and 0.5897 values between 0.58 and 0.8326 values between 0.83 and 1.099 values between 1.09 and 1.3412 values between 1.34 and 1.603 values between 1.60 and 1.854 values between 1.85 and 2.112 values between 2.11 and 2.362 values between 2.36 and 2.622 values between 5.16 and 5.421 values between 6.44 and 6.69offnegativepositiveoften

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. Cam
  2. getC
  3. sem
  4. chos
  5. cid
  6. Gol
  7. rat
  8. uc
  9. cam
  10. cel

Suppresses

  1. &nbsp
  2. StringComparison
  3. Bakanlığı
  4. 끗
  5. 学者
  6. /animations
  7. iado
  8. Autodesk
  9. 经过多年
  10. embros