Neuron Explorer

Qwen3-0.6B

unclear: mostly line breaks, mixed text

A neuron in layer 0 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 and on a mix of things, most often a line break at the end of a line of text. Turning it up made the model write about random topics, such as math or weather, with no clear link to the trigger.

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

    People normally<|endoftext|>Spring Dawn by Meng Haoran 春眠不觉晓 处处闻啼茑 夜来风雨声 花落知多少⏎Pin Yin Version Chun mian bu jue xiao Chu chu wen ti niao

  2. #2 peak

    东方在线 发布人:young⏎14. Energy Resources⏎China is rich in various energy resources. Coal and

  3. #3 peak

    ��性病的⏎anachronism /əˈnækrəˌnɪzə

  4. #4 peak

    Asia, and America. The American variety (var. Americana) is larger. Called⏎whistler, garrot, gowdy, pied widgeon, whiteside, curre, and doucker. Barrow's golden

  5. #5 peak

    ized bed boilers; Circulating fluidized bed boilers⏎- Front<|endoftext|>CNY - 中國元,人民幣⏎最常用 CNY 交叉匯率⏎為什麼您

  6. #6 peak

    , with the introduction of flying boat service to Asia via Pan Am’s famous clippers, Midway became an important refueling and stopover point until war intruded in 1941.⏎Beginning in 1940, facilities

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: 200 The user is a person with a very special way of explaining. The user is a person with a very special way of explaining. I have The user is a person with a very special way of explaining. I

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 others in my awesome Edumatch Voxer group, told me about Break

  2. #2 peak

    soil should be used for building and development,” Bradley explained. “In the whole talk of food production and global warming, soil fertility gets lost in the discourse.”⏎Assistant Professor of Economics Erik Nelson, Bradley’s advisor, says his student’s work is

  3. #3 peak

    rights and freedoms,” National Congress of Australia’s First Peoples co-chair Rod Little said.⏎Assistant Minister for Social Services and Multicultural Affairs Minister Zed Seselja described the

  4. #4 peak

    are never filled. Such shortcomings in medication adherence can decrease the effectiveness of a treatment. This increases the likelihood that

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

None: this is the first layer.

Feeds later layers

  1. L4:1965 weakly excites
  2. L9:232 weakly suppresses
  3. L1:1175 weakly suppresses
  4. L25:1669 weakly suppresses
  5. L6:1281 weakly suppresses
  6. L6:1168 weakly excites
  7. L17:55 weakly excites
  8. L1:699 weakly suppresses
  9. L14:2997 weakly suppresses
  10. L5:1023 weakly suppresses
  11. L16:2560 weakly suppresses
  12. L13:1326 weakly suppresses
  13. L21:351 weakly suppresses
  14. L1:856 weakly excites
  15. L23:1354 weakly suppresses
  16. L23:512 weakly suppresses
  17. L13:885 weakly suppresses
  18. L12:2955 weakly excites
  19. L17:1676 weakly suppresses
  20. L18:583 weakly suppresses
  21. L24:3065 weakly excites
  22. L19:2191 weakly excites
  23. L26:1907 weakly suppresses
  24. L20:2188 weakly suppresses "enough" before "to"

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 -0.24 and -0.231 values between -0.23 and -0.212 values between -0.21 and -0.204 values between -0.20 and -0.197 values between -0.19 and -0.178 values between -0.17 and -0.1615 values between -0.16 and -0.1518 values between -0.15 and -0.1327 values between -0.13 and -0.1235 values between -0.12 and -0.1035 values between -0.10 and -0.0970 values between -0.09 and -0.0898 values between -0.08 and -0.06128 values between -0.06 and -0.05197 values between -0.05 and -0.04309 values between -0.04 and -0.02476 values between -0.02 and -0.01705 values between -0.01 and 0.01464 values between 0.01 and 0.02332 values between 0.02 and 0.03277 values between 0.03 and 0.05195 values between 0.05 and 0.06173 values between 0.06 and 0.08103 values between 0.08 and 0.0975 values between 0.09 and 0.1049 values between 0.10 and 0.1232 values between 0.12 and 0.1334 values between 0.13 and 0.1512 values between 0.15 and 0.169 values between 0.16 and 0.176 values between 0.17 and 0.193 values between 0.19 and 0.204 values between 0.20 and 0.212 values between 0.21 and 0.231 values between 0.28 and 0.301 values between 0.30 and 0.31offnegativepositiveoften

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. oen
  2. amb
  3. 繁
  4. mins
  5. min
  6. .Binding
  7. 分钟左右
  8. cos
  9. mins
  10. len

Suppresses

  1. erate
  2. itore
  3. 屐
  4. ycop
  5. gars
  6. PointerException
  7. estroy
  8. potatoes
  9. xEE
  10. orrar