Neuron Explorer

Qwen3-0.6B

unclear: mid-word subword pieces

A neuron in layer 3 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 a random-looking piece of a longer word, like the 'stein' in a surname or the 'gone' in 'bygone'. When it was turned up, the model's writing drifted toward plain everyday sentences with no 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

    link between blood-vessel health and Alzheimer’s, but this study is “the first one that shows that treating mild cognitive impairment can have results,” says Anton Porsteinsson, MD, director of the Memory Disorders Clinic at the University of Rochester School

  2. #2 peak

    apses, they are merged with the past and become a part of history.⏎Hence, history, in this sense, is the knowledge of the bygone incidents, events, conditions and circumstances of the people in the past. Biographies

  3. #3 peak

    the feel of a bygone era. Time has effectively stood still along these winding country roads, in the small

  4. #4 peak

    parts of the lymphatic system.|⏎|Germ Cell Tumor||GTC. A neoplasm derived from germ cells., normally presenting inside the gonads (ovary and testis).|⏎|Blastoma||More common in

  5. #5 peak

    998).⏎The South Africa Education system, which is grounded on the country’s Constitution, views learners as citizens who can secure a new democracy (Joubert, 2009). Learners’ voices should therefore be heard as

  6. #6 peak

    enteritis. A bacteria called clostridia was isolated as the organism responsible.⏎Necrotizing enteritis is associated with poor hygiene and an inadequate diet, and it was

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: A. The numbers 1, 2, 3, 4, 5 are all related to me, and I need to think about the following question: A. The numbers 1,

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

    2:15|⏎|Last Modified:||11 May 2011 04:

  2. #2 peak

    the largest family is the Eulophidae with nearly 4,500 included species followed closely by the Encyrtidae and Pteromalidae. The phylogenetic relationships between the

  3. #3 peak

    “yarrow” is an Anglo-Saxon word that means “to prepare” because it was

  4. #4 peak

    decrease in pressure, thus a net downward force on the paper.<|endoftext|>On May 28, 1828, began the official life of Kaspar Hauser, a young man who appeared mysteriously in the streets of Nure

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. L2:49 suppresses
  2. L0:1224 suppresses
  3. L0:178 suppresses
  4. L1:87 suppresses
  5. L0:1301 suppresses
  6. L2:166 weakly suppresses
  7. L2:527 weakly excites
  8. L2:2722 weakly suppresses
  9. L1:853 weakly excites
  10. L2:471 weakly suppresses
  11. L0:2565 weakly suppresses
  12. L0:1060 weakly suppresses
  13. L1:1664 weakly suppresses
  14. L0:894 weakly excites
  15. L2:70 weakly excites
  16. L2:1321 weakly suppresses
  17. L1:1029 weakly excites
  18. L2:41 weakly suppresses
  19. L2:2093 weakly suppresses
  20. L1:69 weakly excites
  21. L1:951 weakly excites
  22. L2:1852 weakly excites
  23. L2:641 weakly suppresses
  24. L2:1334 weakly suppresses

Feeds later layers

  1. L7:982 suppresses
  2. L5:1677 suppresses
  3. L10:2623 excites
  4. L4:834 weakly excites
  5. L18:3009 weakly suppresses
  6. L21:764 weakly excites
  7. L18:581 weakly suppresses
  8. L9:2611 weakly excites 'is' right after 'there'
  9. L10:1032 weakly suppresses
  10. L21:1823 weakly suppresses
  11. L13:1879 weakly excites
  12. L6:3013 weakly excites
  13. L11:119 weakly excites
  14. L10:1413 weakly suppresses
  15. L22:695 weakly suppresses
  16. L25:2374 weakly suppresses
  17. L18:2959 weakly excites
  18. L27:91 weakly excites
  19. L23:1005 weakly suppresses
  20. L6:2429 weakly suppresses
  21. L5:504 weakly excites
  22. L22:1039 weakly excites
  23. L16:1297 weakly suppresses unclear: document ends and answer colons
  24. L22:1534 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 -0.46 and -0.433 values between -0.38 and -0.351 values between -0.35 and -0.333 values between -0.33 and -0.304 values between -0.30 and -0.287 values between -0.28 and -0.255 values between -0.25 and -0.2212 values between -0.22 and -0.2019 values between -0.20 and -0.1718 values between -0.17 and -0.1532 values between -0.15 and -0.1246 values between -0.12 and -0.0990 values between -0.09 and -0.07208 values between -0.07 and -0.04456 values between -0.04 and -0.021437 values between -0.02 and 0.01833 values between 0.01 and 0.04326 values between 0.04 and 0.06161 values between 0.06 and 0.09100 values between 0.09 and 0.1253 values between 0.12 and 0.1422 values between 0.14 and 0.1717 values between 0.17 and 0.1916 values between 0.19 and 0.228 values between 0.22 and 0.258 values between 0.25 and 0.275 values between 0.27 and 0.304 values between 0.30 and 0.324 values between 0.32 and 0.352 values between 0.35 and 0.383 values between 0.38 and 0.401 values between 0.40 and 0.431 values between 0.43 and 0.451 values between 0.45 and 0.481 values between 0.56 and 0.58offnegativepositiveoften

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. house
  2. 薨
  3. ply
  4. 享
  5. house
  6. 贵
  7. 下车
  8. ier
  9. 堌
  10. สวย

Suppresses

  1. omorphic
  2. redits
  3. tober
  4. itore
  5. })(
  6. charts
  7. 。",⏎
  8. .refs
  9. IMPORTANT
  10. 舌头