Neuron Explorer

Qwen3-0.6B

zero right after a decimal point

A neuron in layer 18 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 the zero that comes right after a decimal point, as in 0.01 or p < 0.01. It is less active on the digits in citation numbers, which is its weaker side. Turning it up did not change the text in a clear way.

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

    A single heat pulse will switch the magnetic orientation of a nanometer-sized magnetic grain in just a few picoseconds (1 picosecond is 0.000000000001 seconds). As a comparison

  2. #2 peak

    1.176-1.709) and vomiting (p < 0.001, OR = 1.969, 95%

  3. #3 peak

    Meters to millimeters (meter to mm converter) returns numbers of mm giving meters m value. Converting 1 meter to millimeter as 0.001. See also millimeters into meters calculator. See also millimeters into

  4. #4 peak

    3, 95% CI = 1.110-1.529), local citizens (p < 0.001) and officials/clerks (p = 0.00

  5. #5 peak

    ) – this is one thousandth of a meter. In scientific notation, 1 millimeter is 1 – E3 m, which means 0.001 x 1

  6. #6 peak

    .91% of sampled diarrhea patients. The seasonal distribution of norovirus infections was different from non-norovirus patients (p < 0.001), with a half-year peak. Higher proportions of males (p =

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: "If the first person is a " ( " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " " "

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

    the way for all-optical circuits (M Mützel et al 2002 Phys. Rev. Lett. 88 083601-1).⏎The light field of an interference pattern can induce an

  2. #2 peak

    ate the molecular basis of the genetic susceptibility to alcoholism.(JAMA. 1992;268:1877-1882)<|endoftext|>The Imaginative Treatment of Geometry in Waldorf Education⏎A

  3. #3 peak

    , CT scan, rule, minor head injury⏎Ann Afr Surg. 2018;15(2):57-61 DOI:http://dx.doi.org/10.4314

  4. #4 peak

    are and where we need to go in the 21st century.⏎Longwall USA, 1998 Jun; :123-138⏎Although the number of operating longwall mining systems in the U.S.

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. L17:259 excites
  2. L16:211 weakly suppresses
  3. L5:2309 weakly suppresses
  4. L17:1445 weakly excites
  5. L8:2526 weakly suppresses
  6. L9:2524 weakly excites
  7. L11:21 weakly excites
  8. L14:741 weakly suppresses
  9. L14:2796 weakly excites
  10. L17:2502 weakly excites opening parenthesis before an abbreviation
  11. L9:2453 weakly excites
  12. L16:2541 weakly excites
  13. L5:315 weakly suppresses
  14. L5:883 weakly excites
  15. L0:411 weakly suppresses
  16. L9:2401 weakly excites
  17. L12:2528 weakly suppresses digits inside numbers
  18. L7:2137 weakly suppresses
  19. L17:2081 weakly excites
  20. L13:779 weakly excites journal names after 'published in'
  21. L4:2386 weakly excites
  22. L11:1157 weakly excites
  23. L1:2573 weakly suppresses
  24. L9:2849 weakly excites

Feeds later layers

  1. L23:1214 weakly excites
  2. L19:2516 weakly suppresses
  3. L20:2400 weakly suppresses
  4. L20:1246 weakly suppresses
  5. L25:497 weakly suppresses
  6. L26:1633 weakly excites
  7. L23:1381 weakly excites
  8. L19:1653 weakly excites
  9. L25:2174 weakly excites
  10. L25:1100 weakly suppresses
  11. L24:2639 weakly suppresses
  12. L22:1341 weakly excites
  13. L22:449 weakly excites
  14. L19:2668 weakly excites
  15. L22:72 weakly excites
  16. L23:2720 weakly excites
  17. L26:300 weakly suppresses
  18. L19:1136 weakly excites
  19. L26:211 weakly excites
  20. L21:2174 weakly excites
  21. L22:195 weakly excites number right after a year or range word
  22. L19:2590 weakly suppresses
  23. L21:634 weakly excites
  24. L23:2444 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 -5.66 and -5.343 values between -5.03 and -4.721 values between -4.72 and -4.413 values between -4.09 and -3.781 values between -3.78 and -3.472 values between -3.47 and -3.162 values between -2.84 and -2.532 values between -2.22 and -1.901 values between -1.90 and -1.594 values between -1.59 and -1.2810 values between -1.28 and -0.9751 values between -0.97 and -0.65337 values between -0.65 and -0.341402 values between -0.34 and -0.031605 values between -0.03 and 0.29395 values between 0.29 and 0.6038 values between 0.60 and 0.9110 values between 0.91 and 1.2210 values between 1.22 and 1.546 values between 1.54 and 1.856 values between 1.85 and 2.162 values between 2.16 and 2.474 values between 2.47 and 2.793 values between 2.79 and 3.101 values between 3.10 and 3.411 values between 3.41 and 3.731 values between 3.73 and 4.043 values between 4.04 and 4.352 values between 4.66 and 4.981 values between 6.54 and 6.85offnegativepositiveoften

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. NDER
  2. рав
  3. strar
  4. stra
  5. знания
  6. ificance
  7. ật
  8. stringify
  9. berman
  10. kinson

Suppresses

  1. unction
  2. 원
  3. 约束
  4. acket
  5. iov
  6. 膑
  7. 卫
  8. 元
  9. 锟
  10. 次