Neuron Explorer

Qwen3-0.6B

unclear: word before a technical term

A neuron in layer 14 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron tends to fire on the word right before a technical noun, such as 'the' before 'internet' or 'point', or the bracket before 'DC'. There is no clear theme, and turning it up or down did not change the text in any recognisable way.

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

    . There are many styles of diodes. The most common ones in everyday electronics are rectifier diodes, which convert alternating current (AC) to direct current (DC).⏎Verify you are looking at a diode. Diodes

  2. #2 peak

    of Hawai'i at Mānoa and the University of Delaware shined flashes of UV light on their eyes. Photoreceptors, the light-detecting cells

  3. #3 peak

    2)^2 in such a way that its x-coordinate is increasing at a constant rate of 2 units per second. How fast is the particle's distance to the point (-2, 0) changing at the moment that the particle is at

  4. #4 peak

    their visit.⏎This crested souvenir cup was sold by Mrs Emily Yalden who owned a shop selling china and glassware at 17, The Square, Winchester between 1887 and 1914. The cup

  5. #5 peak

    keeps them busy, forming a routine and a sense of responsibility towards them.⏎The companions<|endoftext|>The Internet of Things (IoT) is about connecting any device to the internet and enabling it to transmit data in real time without human intervention. This technology

  6. #6 peak

    decade. - one of the steepest declines experienced by any bird species.⏎In 2003 the culprit was identified - an anti-inflammatory pain-killing drug diclofenac used in veterinary medicine to treat colic in cattle.

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 name of a girl, so the name of the little, and the name of the girl, the name of the girl, the name of the girl, the name of the name of the girl, the name of the name of the

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

    (c. 65) of his only child, a daughter, Flavia Julia, to whom he accorded the title Augusta.⏎Titus, in full Titus Vespasianus Augustus, original name Titus Flavi

  2. #2 peak

    great work Tordesillas made use of all the public archives, having access to documents of every kind. It is evident in<|endoftext|>Vitamin D, sometimes called the sunshine vitamin, plays an important role in a number of your body’s basic functions

  3. #3 peak

    When your cells can’t get glucose (a kind of sugar) to burn for energy, they burn fat instead. This leaves behind acids

  4. #4 peak

    vulnerability on full display and his soldiers’ suffering moved him to tears. It wasn’t affected or a performance; it really did upset the general.⏎Then his long, tall shadow fell across the young Hoosier from northern Indiana. For Private K

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. L12:2419 suppresses
  2. L12:2655 excites
  3. L12:1471 suppresses
  4. L12:1581 excites
  5. L11:1645 suppresses
  6. L12:1436 excites
  7. L10:2264 suppresses
  8. L10:1297 excites
  9. L10:1124 suppresses
  10. L11:2070 excites
  11. L13:1911 suppresses
  12. L13:846 suppresses
  13. L13:1919 excites
  14. L13:2058 excites
  15. L13:1059 suppresses
  16. L1:2366 suppresses
  17. L11:977 excites
  18. L10:1822 excites
  19. L13:110 weakly suppresses
  20. L12:2892 weakly suppresses
  21. L12:840 weakly suppresses
  22. L11:374 weakly suppresses
  23. L11:1345 weakly excites
  24. L13:1323 weakly suppresses

Feeds later layers

  1. L19:2176 suppresses
  2. L18:2235 suppresses
  3. L17:3040 excites
  4. L17:2487 suppresses
  5. L18:529 weakly excites
  6. L17:1298 weakly excites
  7. L16:1742 weakly suppresses
  8. L17:2352 weakly suppresses
  9. L15:617 weakly excites
  10. L21:1994 weakly suppresses
  11. L22:280 weakly suppresses
  12. L15:2340 weakly excites
  13. L16:1108 weakly excites
  14. L20:194 weakly excites
  15. L15:1454 weakly excites
  16. L15:494 weakly excites
  17. L18:2718 weakly suppresses
  18. L15:3047 weakly suppresses preposition introducing a definition or date
  19. L15:1941 weakly excites
  20. L21:1380 weakly excites
  21. L16:1121 weakly suppresses
  22. L22:1839 weakly excites
  23. L18:2790 weakly suppresses
  24. L18:1486 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.

1 values between -2.50 and -2.391 values between -2.19 and -2.081 values between -2.08 and -1.981 values between -1.98 and -1.881 values between -1.77 and -1.672 values between -1.56 and -1.463 values between -1.46 and -1.364 values between -1.36 and -1.2510 values between -1.25 and -1.154 values between -1.15 and -1.058 values between -1.05 and -0.9413 values between -0.94 and -0.849 values between -0.84 and -0.7411 values between -0.74 and -0.6337 values between -0.63 and -0.5360 values between -0.53 and -0.4258 values between -0.42 and -0.32126 values between -0.32 and -0.22258 values between -0.22 and -0.11907 values between -0.11 and -0.011492 values between -0.01 and 0.10469 values between 0.10 and 0.20211 values between 0.20 and 0.3090 values between 0.30 and 0.4143 values between 0.41 and 0.5126 values between 0.51 and 0.6127 values between 0.61 and 0.7211 values between 0.72 and 0.8210 values between 0.82 and 0.934 values between 0.93 and 1.032 values between 1.03 and 1.131 values between 1.13 and 1.246 values between 1.24 and 1.341 values between 1.55 and 1.65offnegativepositiveoften

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. писан
  2. pired
  3. IMPLIED
  4. phạt
  5. ồi
  6. signin
  7. WARRANT
  8. rien
  9. egin
  10. `,⏎

Suppresses

  1. NAME
  2. 姓名
  3. 切れ
  4. 姓
  5. name
  6. 颜色
  7. 名字
  8. aviours
  9. colour
  10. uem