Neuron Explorer

Qwen3-0.6B

opening parenthesis before a citation year

A neuron in layer 8 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 open bracket that comes right after a journal or book title in a reference, just before the year, as in 'Journal of Hydrology (1995)'. It is negative on open brackets in Lisp-style code examples. Turning it up gave text that did not show the citation pattern.

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

    icide and Social Change in China⏎- Cite this article as:⏎- Phillips, M.R., Liu, H. & Zhang, Y. Cult Med Psychiatry (1999) 23: 25. doi:1

  2. #2 peak

    redistribution policy in Brazil."⏎CITATION: Carlos Peres and Maurício Schneider. Subsidized agricultural resettlements as drivers of tropical deforestation. Biological Conservation (2011)<|endoftext|>Neuroimmune Toxicity Recovery Programs⏎Chronic

  3. #3 peak

    groundwater for hydro electric power generation⏎Bardsley, W.E. (1995). Using groundwater for hydro electric power generation. Journal of Hydrology (NZ), 34(1), 1-14.⏎Permanent Research

  4. #4 peak

    . Ben Cohen. "The Racial Geography of the Federal Death Penalty" Washington Law Review (2010).<|endoftext|>Last updated 2/27/15

  5. #5 peak

    T cells. The prominent spikes of the Env protein are visible under hight magnification. (Dewannieux, M. et al., Genome Res. (2006) 16: gr.55657

  6. #6 peak

    005⏎Please cite this report as:⏎Global Volcanism Program, 2007. Report on Macdonald (Undersea Features) (Wunderman, R., ed.). Bulletin of the Global Volcanism Network

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: the power of a mind. When a person is able to understand the meaning of life, they can't help but admire the beauty of it. The beauty of life is... eternal, and it's something that I cannot grasp in my own

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

    ifting has different aims that all are based on psychological motives, and these motives will be mentioned in this article.⏎- Showing love:⏎The theory of “symbolic interactionism”, argues that people communicate through symbols. For example, men buy flowers for

  2. #2 peak

    00) ⇒ #<marker at 100 in markers.texi>⏎;; Now insert one character at the beginning of the buffer: (goto-char (point-min)) ⇒ 1 (insert "Q") ⇒ nil

  3. #3 peak

    its current position.⏎Here are examples of creating markers, setting markers, and moving point to markers:⏎;; Make a new marker that initially does not point anywhere: (setq m1 (make-marker)) ⇒ #<marker in no buffer>⏎m

  4. #4 peak

    and is known only to the Butonian people on the Pulau Makassar. The study used the symbolic interaction theory by Herbert Blumer. The identification of six informers was deemed viable.

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. L3:57 excites
  2. L4:64 excites
  3. L0:734 suppresses
  4. L4:2309 excites
  5. L4:45 weakly suppresses
  6. L5:275 weakly excites
  7. L2:1710 weakly suppresses
  8. L0:1252 weakly suppresses
  9. L1:2470 weakly suppresses
  10. L6:468 weakly excites
  11. L6:1352 weakly excites
  12. L4:1931 weakly suppresses
  13. L6:13 weakly excites
  14. L4:55 weakly suppresses
  15. L0:2639 weakly excites unclear: weak hits on 'holiday' and 'playing', negative on 'Ada-fruit'
  16. L7:122 weakly excites
  17. L4:3066 weakly suppresses
  18. L7:100 weakly suppresses
  19. L5:2970 weakly suppresses
  20. L5:1623 weakly suppresses
  21. L5:1287 weakly suppresses
  22. L1:2605 weakly excites
  23. L6:2219 weakly excites
  24. L2:2242 weakly suppresses

Feeds later layers

  1. L16:1513 excites opening parenthesis before a numbered list
  2. L19:453 excites
  3. L18:682 excites
  4. L10:1041 excites opening parenthesis before phone or code
  5. L9:1414 excites opening parenthesis after copyright or year
  6. L15:2486 excites bracket or parenthesis inside a quote
  7. L22:144 excites
  8. L10:66 excites opening parenthesis after a citation or term
  9. L17:304 excites opening parenthesis after a claim
  10. L17:1509 excites
  11. L12:2803 excites
  12. L12:1443 weakly excites opening parenthesis after a name or term
  13. L26:88 weakly excites
  14. L11:2191 weakly excites unclear: brackets and parentheses in text
  15. L14:2276 weakly excites opening bracket after a term or number
  16. L14:149 weakly suppresses
  17. L12:1180 weakly excites
  18. L18:980 weakly excites opening parenthesis after a noun phrase
  19. L21:1844 weakly excites
  20. L15:324 weakly suppresses
  21. L22:1637 weakly excites
  22. L15:344 weakly excites
  23. L21:1585 weakly excites
  24. L20:722 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.

5 values between -0.56 and -0.4621 values between -0.46 and -0.36129 values between -0.36 and -0.26456 values between -0.26 and -0.16972 values between -0.16 and -0.061298 values between -0.06 and 0.04714 values between 0.04 and 0.14223 values between 0.14 and 0.2458 values between 0.24 and 0.349 values between 0.34 and 0.444 values between 0.44 and 0.543 values between 0.63 and 0.731 values between 0.73 and 0.831 values between 0.83 and 0.931 values between 1.33 and 1.432 values between 1.43 and 1.532 values between 1.53 and 1.631 values between 1.83 and 1.931 values between 2.03 and 2.131 values between 2.13 and 2.231 values between 2.33 and 2.432 values between 2.53 and 2.631 values between 2.93 and 3.031 values between 3.13 and 3.231 values between 3.33 and 3.43offnegativepositiveoften

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. 限り
  2. 截至
  3. stitutions
  4. 降价
  5. _),
  6. assum
  7. نك
  8. ,),⏎
  9. Dist
  10. Dist

Suppresses

  1. チーム
  2. 적으로
  3. igsaw
  4. сек
  5. rel
  6. .Attribute
  7. .Pattern
  8. electrónico
  9. ара
  10. �