Neuron Explorer

Qwen3-0.6B

unclear: word-initial pieces of longer words

A neuron in layer 26 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 the first pieces of longer words that have been split up, such as 'fore' in 'forebears' or 'Fur' in 'Furstenberg'. Turned down, the model got stuck repeating 'perfect' or 'nature'. The pattern is loose and the examples do not share one clear theme.

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

    . Beneath the castle is a network of tunnels, a barracks during the Napoleonic Wars and a military command centre during World War II.<|endoftext|>Our English forebears just partly anglicized a German term back in the sixteenth century

  2. #2 peak

    , la him/her/you formal||los, las them/you plural|⏎A direct object

  3. #3 peak

    A Memoir⏎- The Woman I Wanted to Be. Diane Von Furstenberg⏎- Dressed: A Century of Hollywood Costume Design⏎- Secrets

  4. #4 peak

    (the angle at which a mineral goes black and you rotate the slide) and birefringence (the colour of the mineral using crossed polars).⏎Minerals have different characteristics using the different types of light – plain polarised (P

  5. #5 peak

    along the Pacific Coast. The climate is mild and the soil is fertile and easily farmed.⏎While medieval European medicine was still mired<|endoftext|>ASIAN WILD HORSE⏎SCIENTIFIC NAME: Equus Ferus⏎HABITAT

  6. #6 peak

    sacrifice can be a message for each and every one of us.⏎Our Chamber of Commerce estimates the Treaty Oak to be over 600 years old. Most foresters settle on about 300,

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 fact that a perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect perfect

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

    and beyond.”⏎Yersinia pestis is a bacterium responsible for tens of millions of deaths, most occurring in the course of three devastating global pandemics. Catastrophic as this

  2. #2 peak

    to notice that they are actually different. Mind readers jump to conclusions that are true for them, without checking whether they are true for the other person.⏎5. Catastrophizing: You expect disaster. You notice or hear about a problem and

  3. #3 peak

    scenes?⏎- What cuts are structural.⏎Example: Indiana Jones And The Last Crusade: the boat race in Venice.⏎After facing rats and burning waters in the catacombs<|endoftext|>Google has once again surprised us with another of their now infamous homepage

  4. #4 peak

    caused about $84 billion in losses since then.⏎The Boll Weevil Eradication Program—initiated in 1978 by USDA’s Animal and Plant Health Inspection Service with ARS research support and run by eradication foundations

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. L5:1294 excites
  2. L2:1518 weakly suppresses
  3. L18:420 weakly excites
  4. L18:1023 weakly suppresses the digit 3 inside numbers
  5. L17:134 weakly suppresses
  6. L2:1758 weakly excites
  7. L16:117 weakly suppresses
  8. L2:1622 weakly excites
  9. L17:994 weakly suppresses
  10. L6:4 weakly excites
  11. L17:378 weakly excites
  12. L3:1888 weakly suppresses
  13. L16:14 weakly excites
  14. L9:35 weakly suppresses
  15. L17:385 weakly excites
  16. L4:97 weakly excites
  17. L4:591 weakly excites
  18. L20:147 weakly excites
  19. L16:2808 weakly excites
  20. L10:237 weakly suppresses
  21. L1:6 weakly excites
  22. L22:848 weakly excites
  23. L14:1721 weakly suppresses
  24. L1:23 weakly excites

Feeds later layers

  1. L27:298 weakly excites
  2. L27:2406 weakly excites
  3. L27:2448 weakly excites
  4. L27:1309 weakly excites
  5. L27:2611 weakly suppresses
  6. L27:2556 weakly excites
  7. L27:2039 weakly excites
  8. L27:7 weakly excites
  9. L27:758 weakly excites
  10. L27:1750 weakly suppresses
  11. L27:2056 weakly excites the 'w' in shipwreck
  12. L27:1013 weakly suppresses
  13. L27:2904 weakly excites
  14. L27:1246 weakly suppresses
  15. L27:991 weakly excites
  16. L27:1270 weakly excites
  17. L27:723 weakly excites
  18. L27:1792 weakly suppresses
  19. L27:1726 weakly suppresses
  20. L27:770 weakly excites
  21. L27:1848 weakly suppresses
  22. L27:441 weakly excites
  23. L27:1505 weakly excites
  24. L27:1710 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 -23.31 and -21.861 values between -18.97 and -17.522 values between -13.17 and -11.721 values between -11.72 and -10.275 values between -10.27 and -8.8211 values between -8.82 and -7.3717 values between -7.37 and -5.9241 values between -5.92 and -4.4775 values between -4.47 and -3.02154 values between -3.02 and -1.57983 values between -1.57 and -0.132184 values between -0.13 and 1.32221 values between 1.32 and 2.7781 values between 2.77 and 4.2254 values between 4.22 and 5.6726 values between 5.67 and 7.1216 values between 7.12 and 8.5712 values between 8.57 and 10.027 values between 10.02 and 11.473 values between 11.47 and 12.924 values between 12.92 and 14.373 values between 14.37 and 15.821 values between 15.82 and 17.271 values between 20.16 and 21.612 values between 23.06 and 24.511 values between 24.51 and 25.961 values between 33.21 and 34.66offnegativepositiveoften

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. /LICENSE
  2. ้าง
  3. ื
  4. entrant
  5. roach
  6. uant
  7. ouns
  8. \Queue
  9. �
  10. 眷

Suppresses

  1. CppObject
  2. -parser
  3. 主义者
  4. 总资产
  5. ⁎
  6. ⇑
  7. ActivatedRoute
  8. marvin
  9. ��
  10. Gtk