Neuron Explorer

Qwen3-0.6B

comma opening an inserted clause

A neuron in layer 14 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 a comma in a long sentence, especially the one that sets off a side remark or a new clause, such as "in certain cases, drastic changes". Its steered output did not show a clear theme.

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

    stripe width, color and shape, the distance between the marks and whether it has unique characteristics such as scars or, in the case of some wildebeest, broken or deformed horns.⏎Hundreds of digital images are collected in the field each

  2. #2 peak

    ). In the poorly buffered forest lakes, the decrease in pH was clearly harmful. In clear-water lakes, fish fauna was subject to considerable and, in certain cases, drastic changes. It became obvious that the fish stocks would be<|endoftext|>When we judge

  3. #3 peak

    airport promised to put what had once been a minor railroad town within a day’s reach of London and Paris – and, in the minds of some Atlantans, on the same plane of international significance as those great world cities. But Eugene Black

  4. #4 peak

    electricity in schools, it is recognised that energy for other purposes such as cooking, hot water, and (in some regions) space heating, also plays an important role in improving educational quality.⏎- Education features

  5. #5 peak

    management|⏎Class 10 Science Chapter 4 Environmental management questions answers in Pdf format:⏎Here is what you all are waiting for. Teachers, parents and student, now you will find below the Solutions of Class 10 Science Chapter 4

  6. #6 peak

    technologies were using ultrasound waves or MRI for imaging.⏎However, swept source OCT follows a little bit different mechanism for the same process. Unlike sound waves or MRI techniques, OCT uses light waves for imaging. The shorter wavelength of light empowers this technology

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: I have been thinking about the things I have been thinking about, but I have been thinking about the things I have been thinking about, but I have been thinking about, but I have been thinking about, but I have been thinking about,

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

    be simply due to low testosterone. Men with low testosterone may also have osteopenia or even osteoporosis (thin bones). Testosterone keeps the body skin strong , hairy and with healthy sebaceous glands. Low testosterone can result in a

  2. #2 peak

    addition to the advantages at the chip (LED) level, the nano-membranes (which are thin, at 100nm) can be broken using mechanical force and so the microLEDs can be easily separated from the substrate and

  3. #3 peak

    the Manitowoc River. Many smaller villages or encampments also could be found throughout the area. ...<|endoftext|>The words flour and flower do not only have the same root, but they were also spelled the same

  4. #4 peak

    These skills similarly translate to other areas of their lives, such as dressing ,eating, and in the academic setting.⏎3) Self Regulation– Crafts that require drying require waiting! This is a great lesson for you child to demonstrate self control and patience

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. L12:315 excites
  2. L7:554 suppresses
  3. L10:319 suppresses
  4. L12:876 suppresses
  5. L11:558 excites
  6. L8:395 suppresses
  7. L12:25 excites unclear: mid-phrase words, mixed
  8. L7:1060 suppresses
  9. L10:936 suppresses
  10. L13:414 excites
  11. L5:769 suppresses
  12. L6:2011 suppresses
  13. L9:190 weakly excites
  14. L9:682 weakly suppresses
  15. L8:185 weakly excites
  16. L13:596 weakly excites
  17. L13:1652 weakly suppresses
  18. L11:1759 weakly excites
  19. L9:2286 weakly suppresses
  20. L6:109 weakly suppresses
  21. L5:3 weakly suppresses
  22. L13:2949 weakly excites
  23. L8:86 weakly excites
  24. L11:358 weakly suppresses

Feeds later layers

  1. L16:976 excites
  2. L16:2315 excites
  3. L17:2271 excites
  4. L15:852 weakly excites
  5. L15:2619 weakly excites
  6. L16:1573 weakly excites
  7. L15:2828 weakly excites
  8. L15:3048 weakly excites
  9. L15:2579 weakly excites
  10. L16:2440 weakly excites
  11. L16:491 weakly suppresses
  12. L15:1636 weakly excites
  13. L19:730 weakly suppresses
  14. L15:2298 weakly excites
  15. L16:2211 weakly suppresses
  16. L16:2393 weakly excites
  17. L16:1346 weakly suppresses
  18. L15:2552 weakly excites
  19. L16:433 weakly excites
  20. L15:1505 weakly excites
  21. L27:444 weakly excites
  22. L18:660 weakly excites
  23. L18:1486 weakly suppresses
  24. L18:2470 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 -4.61 and -4.411 values between -4.21 and -4.011 values between -3.81 and -3.613 values between -3.61 and -3.412 values between -3.41 and -3.211 values between -3.21 and -3.013 values between -3.01 and -2.8114 values between -2.81 and -2.609 values between -2.60 and -2.4023 values between -2.40 and -2.2020 values between -2.20 and -2.0028 values between -2.00 and -1.8033 values between -1.80 and -1.6048 values between -1.60 and -1.4069 values between -1.40 and -1.2073 values between -1.20 and -1.00120 values between -1.00 and -0.80164 values between -0.80 and -0.60234 values between -0.60 and -0.40460 values between -0.40 and -0.191096 values between -0.19 and 0.01872 values between 0.01 and 0.21263 values between 0.21 and 0.41126 values between 0.41 and 0.6188 values between 0.61 and 0.8145 values between 0.81 and 1.0140 values between 1.01 and 1.2120 values between 1.21 and 1.4111 values between 1.41 and 1.6111 values between 1.61 and 1.8214 values between 1.82 and 2.023 values between 2.02 and 2.223 values between 2.22 and 2.426 values between 2.42 and 2.621 values between 2.62 and 2.821 values between 2.82 and 3.021 values between 3.22 and 3.42offnegativepositiveoften

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. umber
  2. +B
  3. eson
  4. ummer
  5. ,b
  6. ..."⏎
  7. ainty
  8. らず
  9. "They
  10. yx

Suppresses

  1. 铟
  2. Filed
  3. ène
  4. cele
  5. cele
  6. .Ticks
  7. 直辖
  8. Cele
  9. //{⏎
  10. ��