Neuron Explorer

Qwen3-0.6B

unclear: pieces of medical and Latin-style words

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 weakly on the middle pieces of longer, technical words, such as 'submucosa' and 'oligodendrocytes'. The pattern is faint and the examples do not share a clear theme. Turning it up made the writing drift into unrelated descriptions.

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

    biting. Tail lesions are painful.⏎- Ear lesions: Ear lesions are a collective term for lesions in/at the ears of pigs, but also includes otohematomas, also called blood-ears. Ear

  2. #2 peak

    ). Following the emperor Nero’s death in June 68, Titus was energetic in promoting his father’s candidacy for the imperial crown. Licinius Mucianus, legate of Syria, whom he reconciled with Vespas

  3. #3 peak

    illness and generally does not infect the nervous system. It reactivates later in life in people who have suppressed immune systems. The virus damages oligodendrocytes, the cells in the central nervous system (CNS) which produce my

  4. #4 peak

    wall consists of 4 layers of tissue. From deep (external) to superficial (internal) these are the serosa, muscularis externa, submucosa and mucosa. This layered arrangement follows the same general structure in all regions

  5. #5 peak

    , chief, enteroendocrine cells⏎Lamina propria: connective tissue⏎Muscularis mucosa: two smooth muscle layers⏎|Submucosa||Connective tissue, submucosal (Meissner’s)

  6. #6 peak

    smaller particles using chemical and mechanical digestion.⏎The stomach can perform these roles due to the layers of the stomach wall. These are the gastric mucosa, submucosa, muscularis externa and serosa. All parts of the GI tract

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: 2000000000000000000. I am... I am 20000000000000000000.

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

    D. de Caprona, The Role and Meaning of Structures in Genetic Epistemology. R

  2. #2 peak

    The day after the nationwide general strike was called on October 4, 1934, Lluís Companys, now president of the Generalitat, proclaimed the "Republic of Catalunya within the Federal Republic of Spain"—an

  3. #3 peak

    to cross the bridge just before the long tunnel. Source: Facebook⏎The efforts of those who managed the huge task of building the railway through the Sierra de Alfàbia – part of the substantial Sierra de Tramuntana mountains on the north

  4. #4 peak

    studies in more clinically relevant models. Proceedings of the National Academy of Sciences USA (PNAS) has published the study online this week.⏎Argentinian scientist Andrés Dekanty, a Juan de la Cierva researcher in the Mil

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. L1:2677 weakly suppresses
  2. L2:2915 weakly suppresses
  3. L2:24 weakly excites
  4. L1:651 weakly suppresses
  5. L0:2537 weakly suppresses
  6. L1:2022 weakly excites
  7. L2:2601 weakly suppresses
  8. L1:2576 weakly suppresses
  9. L1:1688 weakly excites
  10. L2:2855 weakly suppresses
  11. L1:835 weakly suppresses
  12. L2:2737 weakly suppresses
  13. L2:2432 weakly excites
  14. L1:198 weakly suppresses
  15. L0:222 weakly excites
  16. L1:2937 weakly suppresses
  17. L2:833 weakly suppresses
  18. L0:1149 weakly suppresses
  19. L1:2378 weakly suppresses
  20. L0:253 weakly suppresses
  21. L1:1130 weakly suppresses
  22. L0:2664 weakly suppresses
  23. L1:488 weakly suppresses 'of' after group or combination nouns
  24. L1:685 weakly excites

Feeds later layers

  1. L5:30 suppresses
  2. L4:1473 weakly suppresses
  3. L7:2446 weakly suppresses
  4. L5:10 weakly suppresses
  5. L5:2110 weakly suppresses
  6. L19:2180 weakly excites
  7. L26:2656 weakly suppresses
  8. L6:2952 weakly suppresses
  9. L12:1416 weakly suppresses
  10. L9:735 weakly suppresses
  11. L17:1854 weakly excites
  12. L14:728 weakly excites
  13. L22:2328 weakly excites
  14. L6:2027 weakly excites
  15. L18:1001 weakly excites
  16. L6:1723 weakly suppresses
  17. L20:103 weakly suppresses
  18. L4:1888 weakly suppresses
  19. L6:69 weakly excites
  20. L5:1142 weakly suppresses
  21. L26:318 weakly excites
  22. L4:1543 weakly suppresses
  23. L4:1265 weakly suppresses
  24. L20:658 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 -1.61 and -1.541 values between -0.65 and -0.582 values between -0.51 and -0.447 values between -0.44 and -0.3715 values between -0.37 and -0.3053 values between -0.30 and -0.23132 values between -0.23 and -0.17375 values between -0.17 and -0.10719 values between -0.10 and -0.031209 values between -0.03 and 0.04848 values between 0.04 and 0.11366 values between 0.11 and 0.18119 values between 0.18 and 0.2541 values between 0.25 and 0.3114 values between 0.31 and 0.381 values between 0.38 and 0.452 values between 0.52 and 0.592 values between 0.73 and 0.801 values between 1.07 and 1.14offnegativepositiveoften

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. urma
  2. держива
  3. abay
  4. alom
  5. ecause
  6. alpha
  7. suggest
  8. лей
  9. eden
  10. raud

Suppresses

  1. etical
  2. 化进程
  3. reira
  4. gien
  5. ernity
  6. 庆典
  7. étique
  8. aunch
  9. öff
  10. Launching