Neuron Explorer

Qwen3-0.6B

the word 'called' after 'is' or 'are'

A neuron in layer 15 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 word 'called' when it introduces the name of a thing, as in 'it is called apoptosis'. It is a weak, mostly negative signal on words like 'tissues' or 'infection' that appear right before a noun ends. Turning it up made the model write in a looser, list-like style with repeated 'and' phrases.

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

    opportunity to name an asteroid from which an upcoming NASA mission will return the first samples to Earth.⏎Scheduled to launch in 2016, the mission is called the Origins-Spectral Interpretation-Resource Identification-Security-Regolith Explorer (

  2. #2 peak

    one of those them- then you might not want to read about this giant new rat species which was discovered deep in the jungle of Papua New Guinea.⏎It is called the “Bosavi Woolly Rat” and was discovered during the filming of

  3. #3 peak

    oning but make sure to consult a calendar for exact dates if you are making plans.⏎There are spring parades that occur some three weeks before Easter. These are called Sommertagszug (Summer Day Parade) and are basically a time when

  4. #4 peak

    become real property. Personal property converts to real property when it is permanently, immovably attached to other real property.⏎The conversion from personal to real property is called annexation. Annexation occurs through installing, fastening

  5. #5 peak

    between the battery banks to the panel array. The battery bank will then be disconnected, which regulates the battery voltage and makes it constant. This quick switch method is called PWM, and what this does is it makes sure that your battery bank is charged

  6. #6 peak

    ritual manuals used by priests engaged in the performance of the grander Vedic sacrifices, those requiring three fires and the services of many specialized priests. The manuals are called shrauta (from Sanskrit shruti, “revelation”; literally

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 story of the lives of the people who who have been created by the gods and the lives of the people who who have been created by the and the, and, and, and, and, and, and, and, 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

    Tear, it is a must to know what a tendon is and what are the common areas where a Tendon Tear can occur. Tendons are fibrous tissues which attach a muscle to a bone in the

  2. #2 peak

    to date on all their necessary shots. Here's a summary of the different vaccines to consider.⏎• Flu vaccine: This shot protects against influenza, a serious viral infection that affects many Americans each year. Anyone who wants to reduce his or her chance

  3. #3 peak

    components of the bloodstream, coordinating with platelet cells to stop bleeding at the site of an injury by forming into a gel-like plug. Platelets are normal cell fragments that originate in the bone marrow and travel in the blood as it circulates through

  4. #4 peak

    the formation of spermatozoa and ova, the white blood cells help with protecting the body against infections, the somatic cells are the undifferentiated cells which help the body grow.Goblet cells are present in our body for performing

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. L14:1762 excites
  2. L14:1354 suppresses
  3. L13:1027 excites
  4. L14:361 suppresses
  5. L13:3008 weakly excites
  6. L14:1812 weakly excites
  7. L13:2406 weakly suppresses
  8. L14:2707 weakly excites
  9. L1:785 weakly excites
  10. L7:2796 weakly suppresses
  11. L14:2507 weakly excites
  12. L9:865 weakly suppresses
  13. L14:2137 weakly excites
  14. L14:53 weakly excites
  15. L14:2587 weakly suppresses
  16. L14:673 weakly suppresses
  17. L14:344 weakly excites
  18. L13:1828 weakly suppresses
  19. L14:2181 weakly excites unclear: weak, varied mid-sentence words
  20. L13:612 weakly suppresses
  21. L7:208 weakly excites
  22. L13:1097 weakly excites
  23. L13:786 weakly excites
  24. L14:1114 weakly excites

Feeds later layers

  1. L16:357 strongly suppresses
  2. L18:1655 strongly excites 'and' joining two clauses
  3. L21:739 excites 'both X and Y' pairing words
  4. L24:205 excites
  5. L17:2911 excites
  6. L16:2738 excites
  7. L20:2392 excites
  8. L23:379 excites
  9. L17:1856 suppresses
  10. L17:177 suppresses
  11. L26:65 suppresses
  12. L16:3060 excites
  13. L27:289 weakly excites
  14. L19:612 weakly suppresses
  15. L22:162 weakly excites
  16. L24:1428 weakly suppresses
  17. L18:2663 weakly excites comma ending a clause before an action
  18. L16:331 weakly suppresses
  19. L20:1043 weakly suppresses
  20. L24:140 weakly suppresses
  21. L22:632 weakly excites
  22. L26:586 weakly suppresses
  23. L17:1099 weakly excites
  24. L17:2221 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 -3.45 and -3.302 values between -3.14 and -2.991 values between -2.84 and -2.691 values between -2.39 and -2.241 values between -1.93 and -1.781 values between -1.63 and -1.482 values between -1.18 and -1.021 values between -1.02 and -0.878 values between -0.87 and -0.7214 values between -0.72 and -0.5745 values between -0.57 and -0.42204 values between -0.42 and -0.27568 values between -0.27 and -0.111591 values between -0.11 and 0.04933 values between 0.04 and 0.19351 values between 0.19 and 0.34110 values between 0.34 and 0.4935 values between 0.49 and 0.6415 values between 0.64 and 0.804 values between 0.80 and 0.957 values between 0.95 and 1.101 values between 1.10 and 1.252 values between 1.25 and 1.401 values between 1.40 and 1.553 values between 1.55 and 1.701 values between 1.70 and 1.863 values between 1.86 and 2.011 values between 2.16 and 2.311 values between 2.46 and 2.61offnegativepositiveoften

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. aland
  2. and
  3. 并且
  4. 并
  5. 并于
  6. 和
  7. And
  8. -and
  9. và
  10. և

Suppresses

  1. :+:
  2. that
  3. 弟子
  4. amed
  5. 栎
  6. veled
  7. apest
  8. אים
  9. atsby
  10. velt