Neuron Explorer

Qwen3-0.6B

space after a dash in number ranges

A neuron in layer 0 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 lights up on the space right after a dash or plus sign that joins two numbers or dates, such as '1907 - 1954' or 'Published - 2014'. Turning it up made the model output long strings of zeros and repeated numbers.

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

    Trial (19 April - 22 December 1947)⏎- The IG Farben Trial (

  2. #2 peak

    is expected to increase, which is a growing public health concern.⏎|Status||Published - 2014 jan. 1|⏎- Folkhälsovetens

  3. #3 peak

    ||173|⏎|ISBN (Print)||9 781742 589985|⏎|Publication status||Published - 2015|<|endoftext|>writing essay shiz⏎Word count: 3

  4. #4 peak

    in highly trained racehorses.⏎|Number of pages||7|⏎|Journal||Veterinarija ir Zootechnika|⏎|Publication status||Published - 17 Oct 2008|⏎- age-related changes⏎- equ

  5. #5 peak

    lo?⏎Frida Kahlo (1907 - 1954) was a Mexican artist born in 1907

  6. #6 peak

    different card types in this set:⏎64 Question Card + 16 Playing Cards)⏎- Wh Question Cards (Christmas theme)⏎- What am

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 past, the present, and the future. I have been thinking about the past, the present, and the future. I have been thinking about the past, the present, and the future. I have been

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

    to the animals.⏎In the last quarter ofthe 20th century, maize replaced sorghum as the most important cereal

  2. #2 peak

    (S.W.), the New England Confederation and the Fur Trade Dutch/French (N.Y. &N.F.) is the third supporting idea. * Amerindians forced to do unpaid

  3. #3 peak

    majesty.⏎The third division (verses 1Ch 16:34-36 ) is a quotation of the opening and closing sentences ofPsalms 106:1-48; Psal

  4. #4 peak

    5 Y 2 . Solution : X =- 5 W c) Find some solution of A X = 3 Y

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. L2:501 suppresses
  2. L15:2245 weakly suppresses spot right before a number
  3. L18:99 weakly suppresses
  4. L26:1864 weakly excites
  5. L19:1982 weakly excites
  6. L4:2933 weakly excites
  7. L10:2853 weakly suppresses
  8. L14:1273 weakly suppresses
  9. L27:416 weakly excites
  10. L7:1939 weakly excites
  11. L2:63 weakly excites
  12. L10:2066 weakly suppresses
  13. L11:1477 weakly suppresses
  14. L5:1252 weakly excites
  15. L12:1307 weakly suppresses
  16. L3:1340 weakly suppresses
  17. L9:21 weakly suppresses
  18. L15:2847 weakly excites
  19. L13:1097 weakly excites
  20. L9:1354 weakly excites
  21. L25:1827 weakly suppresses
  22. L13:2309 weakly excites
  23. L27:463 weakly excites
  24. L9:403 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 -0.38 and -0.362 values between -0.31 and -0.296 values between -0.29 and -0.271 values between -0.27 and -0.256 values between -0.25 and -0.225 values between -0.22 and -0.2014 values between -0.20 and -0.189 values between -0.18 and -0.1627 values between -0.16 and -0.1336 values between -0.13 and -0.1178 values between -0.11 and -0.09137 values between -0.09 and -0.07243 values between -0.07 and -0.04432 values between -0.04 and -0.021010 values between -0.02 and 0.00838 values between 0.00 and 0.03457 values between 0.03 and 0.05251 values between 0.05 and 0.07144 values between 0.07 and 0.0973 values between 0.09 and 0.1270 values between 0.12 and 0.1427 values between 0.14 and 0.1613 values between 0.16 and 0.1813 values between 0.18 and 0.217 values between 0.23 and 0.251 values between 0.25 and 0.271 values between 0.27 and 0.301 values between 0.32 and 0.341 values between 0.36 and 0.392 values between 0.43 and 0.462 values between 0.50 and 0.52offnegativepositiveoften

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. пит
  2. :host
  3. mixing
  4. mixes
  5. iddles
  6. wave
  7. ㄇ
  8. 广泛应用
  9. queeze
  10. waves

Suppresses

  1. avors
  2. Larson
  3. bih
  4. away
  5. even
  6. itore
  7. lander
  8. ту
  9. ivant
  10. atie