Neuron Explorer

Qwen3-0.6B

ampersand before last author in citations

A neuron in layer 13 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 "&" or "and" that joins the last two author names in a academic reference, like "Pyszczynski, Solomon, & Greenberg". It is mostly quiet elsewhere. Turning it up gave rambling text that did not clearly connect 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

    helps improve writing competence.⏎Dyson, A. (2008). Playing with textual toys. In Flood, J., Heath, S.B., & Lapp, D. (Eds.), Handbook of research on teaching literacy through

  2. #2 peak

    L., Allison, A. & James, R. 2016. Dorcopsis atrata.

  3. #3 peak

    addus VC, Light RW. Pleural effusion. In: Mason RJ, Murray JF, Broaddus VC, et al., eds. Murray and Nadel's Textbook of Respiratory Medicine. 5th ed.

  4. #4 peak

    ifying fears, and creating dependency. Reminding people of their mortality tends to make them hold more closely to traditional culture (Pyszczynski, Solomon, & Greenberg, 2003); this has implications for mental health,

  5. #5 peak

    man & Horowitz, 2003). Studies find that people do not treat computers as people outside the mediation context (Goldstein, Alsio, & Werdenhoff, 2002) – just as people do not

  6. #6 peak

    , 2018, the Trump administration reversed the policy, just eight days after the photo of Yanela was taken (Shear, Goodnough & Haberman

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 think I have to I have I have I have I have I have I have I have I have I have I have i have i i i i i i i i i i i i i i i i i i i i 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

    “We Were Soldiers Once . . . And Young,” by Lt. Gen. Harold G. Moore and war correspondent Joseph G

  2. #2 peak

    to read. Reading is fascinating. It's full of surprises. And it will take you places you can't get to in any other way.⏎-

  3. #3 peak

    :<|endoftext|>The commerce of the ancient world was gathered into the warehouses of Tyre. "Tyrian merchants were the first who ventured to navigate the Mediterranean waters; and they founded their colonies on the coasts and neighbouring islands of the Aegean

  4. #4 peak

    see the Glossary.<|endoftext|>Madison is preparing for battle against the emerald ash borer, a beetle that has laid waste to millions of ash trees. And city officials are asking for residents' help.⏎A native of eastern Asia, the

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. L4:57 suppresses
  2. L9:92 suppresses "and" or "or" joining two items
  3. L5:548 excites
  4. L10:74 excites
  5. L12:51 excites
  6. L1:1250 weakly suppresses
  7. L0:313 weakly excites "and" joining paired nouns
  8. L9:1408 weakly suppresses
  9. L6:622 weakly excites
  10. L11:1457 weakly suppresses
  11. L6:527 weakly suppresses
  12. L1:1710 weakly suppresses
  13. L0:2266 weakly excites
  14. L10:1389 weakly excites
  15. L0:1656 weakly suppresses
  16. L6:1955 weakly suppresses
  17. L6:792 weakly suppresses
  18. L9:619 weakly excites
  19. L7:2778 weakly suppresses
  20. L11:1917 weakly excites
  21. L9:699 weakly excites
  22. L6:1313 weakly excites
  23. L3:2973 weakly suppresses
  24. L10:2432 weakly excites

Feeds later layers

  1. L16:2442 excites
  2. L17:2766 weakly suppresses
  3. L14:2054 weakly suppresses
  4. L14:1116 weakly excites 'and' inside lists
  5. L17:727 weakly suppresses
  6. L16:2049 weakly suppresses
  7. L17:3021 weakly suppresses
  8. L23:96 weakly excites
  9. L16:980 weakly suppresses
  10. L21:639 weakly suppresses
  11. L27:2512 weakly excites
  12. L16:376 weakly suppresses 'or' joining alternatives
  13. L22:1157 weakly excites
  14. L20:964 weakly suppresses
  15. L17:1450 weakly excites
  16. L18:1755 weakly excites
  17. L19:408 weakly excites
  18. L20:2360 weakly suppresses
  19. L17:293 weakly suppresses
  20. L15:566 weakly suppresses
  21. L17:1085 weakly excites
  22. L18:1625 weakly suppresses
  23. L19:955 weakly suppresses
  24. L15:3011 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.

2 values between -3.55 and -3.391 values between -3.07 and -2.911 values between -2.91 and -2.761 values between -2.76 and -2.602 values between -2.60 and -2.442 values between -2.12 and -1.971 values between -1.97 and -1.813 values between -1.81 and -1.6510 values between -1.65 and -1.497 values between -1.49 and -1.345 values between -1.34 and -1.186 values between -1.18 and -1.027 values between -1.02 and -0.869 values between -0.86 and -0.708 values between -0.70 and -0.5522 values between -0.55 and -0.3973 values between -0.39 and -0.23524 values between -0.23 and -0.071805 values between -0.07 and 0.091050 values between 0.09 and 0.24302 values between 0.24 and 0.4050 values between 0.40 and 0.566 values between 0.56 and 0.721 values between 0.72 and 0.884 values between 0.88 and 1.033 values between 1.03 and 1.191 values between 1.19 and 1.351 values between 1.98 and 2.141 values between 2.61 and 2.77offnegativepositiveoften

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. gable
  2. oz
  3. oir
  4. o
  5. entieth
  6. icky
  7. pter
  8. ]
  9. respects
  10. respect

Suppresses

  1. 不管
  2. Ethics
  3. 笼
  4. NewLabel
  5. ادات
  6. 星座
  7. alot
  8. роб
  9. traî
  10. lems