Neuron Explorer

Qwen3-0.6B

second digit of a year range

A neuron in layer 13 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 the last digit of a year written as a span, such as '2011-2012' or '2016-2017', where the model is filling in the end year. Turning it up gave random text, so the steering does not show this pattern, but the examples are very consistent.

Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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

    BUILDINGS.<|endoftext|>"The U.S. Food and Drug Administration announced today that it has approved the influenza vaccine formulation for the 2011-2012 vaccine that will be used by the six manufacturers licensed to produce and distribute influenza

  2. #2 peak

    programs like The Bachelor, Big Brother, and Survivor that were among the 100 highest-rated youth programs for the 2012-2013 TV season.⏎“E-cigarette companies advertise to a broad TV audience that

  3. #3 peak

    equipment⏎Teacher role modeling of all lessons⏎Core curriculum content integration⏎Physical fitness and nutrition integration⏎Our lessons for the 2012-2013 school year will reflect a physical education program that stresses physical fitness, nutrition,

  4. #4 peak

    17-18 Extension Wheat Production Guide, Georgia farmers planted 160,000 acres in the 2016-2017 season, which marked the third year the state’s wheat acreage declined.

  5. #5 peak

    be kind of funny, but it is 104 times all the coffee that was produced in the world in the 2016-2017 harvest

  6. #6 peak

    other changes and updates, the city is better equipped to handle the kind of emergencies that can devastate a city. 2011 and 2012 saw a long string of severe weather disasters that killed hundreds and destroyed billions in

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: that the role of the community in the role of the family in the role of the social in the role of the community in the role of the role of the role of the role of the role of the role of the role of the role

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

    The black ribbon symbolizes the mourning of a loss, and the blue ribbon represents police officers who have sworn to protect and serve the citizens of their communities.⏎

  2. #2 peak

    closed doors and is elitist, giving rise to conspiracies about who produced the evidence and for what purpose. In this environment, subjectivity prevails over objectivity as policy-makers cherry-pick the evidence to fit with the preconceived views

  3. #3 peak

    between the voltage and current, it is known as the active power. The second term is time-varying, it depends<|endoftext|>The Difference Between LDL and HDL Cholesterol: Blood cholesterol is an important indicator of cardiovascular health. Maintaining optimal

  4. #4 peak

    old scientific infighting.⏎Planetary Protection has two basic missions. “Backward protection” means making sure that nothing dangerous in space gets back to Earth. “Forward protection” is making sure that no Earthly gunk makes it off-world—

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. L9:83 strongly suppresses
  2. L12:1813 strongly excites
  3. L12:1097 strongly excites
  4. L12:1753 suppresses
  5. L9:412 suppresses
  6. L12:1258 excites
  7. L12:2383 suppresses
  8. L6:363 excites
  9. L11:530 suppresses
  10. L11:2122 excites
  11. L6:41 excites
  12. L12:198 suppresses
  13. L12:1682 excites
  14. L12:1783 suppresses
  15. L12:863 excites
  16. L9:386 suppresses unclear: pieces inside numbers and blanks
  17. L7:222 suppresses
  18. L11:47 suppresses
  19. L12:316 suppresses
  20. L7:85 excites
  21. L10:1185 excites
  22. L12:2843 excites
  23. L7:309 excites
  24. L6:955 suppresses

Feeds later layers

  1. L15:1494 suppresses
  2. L15:483 excites
  3. L14:1010 suppresses
  4. L15:250 excites
  5. L14:1210 suppresses
  6. L14:1201 suppresses
  7. L15:1037 suppresses
  8. L15:751 suppresses
  9. L15:3050 excites
  10. L18:2015 suppresses
  11. L15:1265 suppresses
  12. L14:1886 excites
  13. L14:1813 suppresses
  14. L14:2521 suppresses
  15. L14:1089 weakly suppresses
  16. L14:678 weakly suppresses
  17. L15:2279 weakly suppresses
  18. L14:858 weakly suppresses
  19. L14:207 weakly excites
  20. L15:454 weakly suppresses
  21. L15:1173 weakly suppresses
  22. L14:2845 weakly suppresses
  23. L17:1569 weakly suppresses
  24. L16:2045 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.

2 values between -8.87 and -8.601 values between -7.25 and -6.981 values between -6.98 and -6.721 values between -6.72 and -6.451 values between -5.91 and -5.641 values between -5.37 and -5.102 values between -4.56 and -4.292 values between -4.03 and -3.761 values between -3.76 and -3.496 values between -3.49 and -3.224 values between -3.22 and -2.953 values between -2.95 and -2.688 values between -2.68 and -2.4114 values between -2.41 and -2.1413 values between -2.14 and -1.8719 values between -1.87 and -1.6031 values between -1.60 and -1.3423 values between -1.34 and -1.0762 values between -1.07 and -0.8085 values between -0.80 and -0.53198 values between -0.53 and -0.26997 values between -0.26 and 0.011769 values between 0.01 and 0.28576 values between 0.28 and 0.5582 values between 0.55 and 0.824 values between 0.82 and 1.091 values between 1.09 and 1.361 values between 1.62 and 1.89offnegativepositiveoften

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. adt
  2. ulta
  3. alling
  4. 」⏎⏎
  5. inner
  6. acular
  7. adel
  8. */⏎
  9. 」「
  10. 门槛

Suppresses

  1. oose
  2. ntohs
  3. estruct
  4. societies
  5. _FUNCTION
  6. త
  7. 的社会
  8. societal
  9. Ọ
  10. archical