Neuron Explorer

Qwen3-0.6B

numbers inside citation references

A neuron in layer 11 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

Fires on the volume or issue number in a reference, like the 3 in a journal entry such as 2008;309. It stays quiet on most other numbers, and when it is turned up the writing drifts into unrelated topics without any clear link to citations.

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

    . 2008. Do crocodilians get the flu? Looking for influenza A in captive crocodilians. Journal of Experimental Zoology. 309A:1-10. Interpretive Summary: Since crocod

  2. #2 peak

    be well represented in the efforts currently under way to elucidate the molecular basis of the genetic susceptibility to alcoholism.(JAMA. 1992;268:1877-1882)<|endoftext|>The Imag

  3. #3 peak

    ar abscess, mastoiditis, and rheumatic fever in children: time trend analysis. BMJ. August 6, 2005;331:328–9.⏎Used with permission from

  4. #4 peak

    was in the midst of the ‘War<|endoftext|>According to a recent study from the Journal of the American Medical Association (JAMA. 2010;303(18):1848-1

  5. #5 peak

    Am J Roentgenol 2008;190:273.) This traditionally is focused on reducing unnecessary CT scans

  6. #6 peak

    role in setting up or maintaining the 180 degrees rotational symmetry of the two flagella.⏎Rights and Permissions⏎Citation: J Cell Biol. 1983 Sep;97(3):9

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 can't be remembered. I can be remembered. I can be heard. I can be seen. I can be seen. I can be seen. I can be heard. I can be remembered. I can be seen. I can

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

    weeks old. Then, check the back teeth; if it has yellowish stains/tartar buildup, your cat is likely around 1 to 2 years old.⏎It doesn’t end there; now, check for tartar buildup

  2. #2 peak

    ⏎7 letter words⏎6 letter words⏎5 letter words⏎4 letter words⏎3 letter words⏎2 letter words⏎Images for ZEATINSLoading...⏎SCRAB

  3. #3 peak

    ask you to walk on our treadmill for up to 15 minutes. You are free to rest at any time. The whole process may take between one and two hours.⏎What is the most important piece of technology used in the

  4. #4 peak

    Wang Bei (2008). New Insight Into Mencius' Theory of the Original Goodness in Human Nature. Frontiers of Philosophy in China 3 (1):27 - 38.⏎Katrin Froese (2

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. L10:2354 suppresses
  2. L10:1602 excites the number 1 as a unit value
  3. L4:205 weakly excites the digit 2 inside citations
  4. L4:2474 weakly excites the word 'constant' in science and maths
  5. L5:746 weakly excites
  6. L9:1348 weakly suppresses
  7. L10:726 weakly excites
  8. L9:2840 weakly suppresses
  9. L9:977 weakly excites the last digit of a decade year
  10. L6:227 weakly suppresses
  11. L6:1286 weakly suppresses
  12. L0:646 weakly excites
  13. L5:166 weakly excites
  14. L3:119 weakly suppresses
  15. L8:2269 weakly suppresses
  16. L1:2220 weakly excites
  17. L6:980 weakly suppresses
  18. L7:683 weakly suppresses
  19. L8:184 weakly suppresses
  20. L5:1878 weakly suppresses
  21. L4:1886 weakly excites
  22. L9:1032 weakly suppresses
  23. L4:1470 weakly excites
  24. L7:844 weakly suppresses

Feeds later layers

  1. L12:1272 suppresses
  2. L13:1397 suppresses
  3. L18:1622 weakly suppresses
  4. L16:1120 weakly excites
  5. L12:947 weakly excites
  6. L15:1758 weakly suppresses
  7. L12:455 weakly excites
  8. L18:2895 weakly suppresses
  9. L25:694 weakly excites
  10. L19:2546 weakly excites
  11. L20:2979 weakly excites
  12. L18:1264 weakly excites
  13. L16:1060 weakly suppresses
  14. L20:2234 weakly excites
  15. L27:2027 weakly excites
  16. L13:2276 weakly suppresses
  17. L24:1852 weakly excites
  18. L17:454 weakly excites
  19. L13:2098 weakly suppresses
  20. L20:2667 weakly excites
  21. L18:2221 weakly excites
  22. L19:1487 weakly suppresses
  23. L17:2189 weakly suppresses
  24. L19:1312 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 -3.87 and -3.741 values between -2.31 and -2.181 values between -2.18 and -2.051 values between -1.79 and -1.662 values between -1.40 and -1.272 values between -1.27 and -1.141 values between -1.14 and -1.021 values between -1.02 and -0.893 values between -0.89 and -0.763 values between -0.76 and -0.633 values between -0.63 and -0.5026 values between -0.50 and -0.37117 values between -0.37 and -0.24579 values between -0.24 and -0.111578 values between -0.11 and 0.031157 values between 0.03 and 0.16336 values between 0.16 and 0.2977 values between 0.29 and 0.4211 values between 0.42 and 0.553 values between 0.55 and 0.681 values between 0.80 and 0.932 values between 0.93 and 1.061 values between 1.06 and 1.191 values between 1.19 and 1.32offnegativepositiveoften

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. isans
  2. ilia
  3. ISION
  4. цов
  5. ок
  6. _mp
  7. urpose
  8. inity
  9. สา
  10. isas

Suppresses

  1. exter
  2. 回来了
  3. cci
  4. ster
  5. decor
  6. letter
  7. GetX
  8. ativo
  9. 图书
  10. 回来