Neuron Explorer

Qwen3-0.6B

opening parenthesis after a title or name

A neuron in layer 12 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 an opening bracket that follows a title, a name or a study label, like the '(' in 'Friends (NBC)' or a citation. Turning it up made the writing wander without a clear topic.

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

    no sex, or no television. Afterwards, the students answered questions about their sexual attitudes and beliefs.⏎Taylor used video clips from six television programs: "Friends" (NBC), "Just Shoot Me" (NBC), "Ally McBeal

  2. #2 peak

    the philosophy of Plato and the strong defense of Aristotle expressed in several works, especially his Comparatio Platonis et Aristotelis / Comparison of Plato and Aristotle (1458). This book accused the humanist circle patronized by the

  3. #3 peak

    after first sex, the great majority of teens don’t," says Meier.⏎Meier’s study, “Adolescent Sexual Activity and Well- Being" (funded in part by the National Institute of Child Health and Development), examined psychological

  4. #4 peak

    Flood, J., Heath, S.B., & Lapp, D. (Eds.), Handbook of research on teaching literacy through the communicative and visual arts (Vol 2, pp. 4

  5. #5 peak

    in stigma attached to the term intellectual disability in comparison with the previous term, mental retardation. An online questionnaire, based on the Multidimensional Attitudes Scale (Findler, Vilchinsky, & Werner, 2007),

  6. #6 peak

    of Science degree in Computer Science. This program is growing to meet the national demand for CS majors, and has two major concentrations: A. programming and software engineering (we teach Python, assembly, C, C++, Java, Matlab, VHDL

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: it's a list of common questions and a list of resources for the latest updates on the Internet is a list of references for the "Sino-English" "Baidu Baike" and "Baidu Baike" is the website

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

    00 wounded.⏎- Richter, G (1870). Manual of Coorg. Stolz. p. 337. Retrieved 14 August 2014.⏎- Archer, Jeremy. "Sir

  2. #2 peak

    2-04-15. http://web.mit.edu/newsoffice/1992/citation-0415.html. Retrieved 200<|endoftext|>Our herd of awesome goats clears land and

  3. #3 peak

    grams (helpful to know if you're counting carbs).⏎Protein - For Protein, one serving of this food has a total of 1 gram.⏎Minerals - In

  4. #4 peak

    serving size available.⏎- These nutrition facts came directly from the USDA or manufacturer/restaurant.⏎- If you're using a calorie counter, remember that Fat, Carbs, and Protein calories are just close estimations based on the Atwater factors:⏎

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. L10:271 weakly excites
  2. L4:701 weakly suppresses
  3. L6:1389 weakly excites
  4. L9:20 weakly suppresses
  5. L7:2684 weakly suppresses
  6. L8:305 weakly excites
  7. L9:188 weakly excites 'of' in 'National Institute of' names
  8. L11:1019 weakly suppresses
  9. L11:785 weakly suppresses
  10. L9:2406 weakly suppresses
  11. L6:350 weakly suppresses
  12. L11:216 weakly excites
  13. L2:1519 weakly suppresses
  14. L3:1421 weakly suppresses
  15. L10:1225 weakly excites
  16. L10:1039 weakly suppresses
  17. L9:341 weakly suppresses
  18. L0:1962 weakly suppresses
  19. L7:343 weakly excites
  20. L6:990 weakly excites
  21. L9:1414 weakly suppresses opening parenthesis after copyright or year
  22. L4:1428 weakly excites
  23. L7:1315 weakly excites
  24. L8:1469 weakly suppresses

Feeds later layers

  1. L13:1324 weakly excites
  2. L15:1024 weakly suppresses
  3. L15:2407 weakly excites
  4. L13:303 weakly suppresses
  5. L13:1148 weakly excites
  6. L18:2253 weakly excites
  7. L16:1039 weakly suppresses
  8. L15:498 weakly suppresses
  9. L17:3005 weakly suppresses
  10. L16:1131 weakly excites
  11. L27:1954 weakly suppresses
  12. L14:1963 weakly excites
  13. L15:2922 weakly excites
  14. L21:2041 weakly suppresses
  15. L18:2912 weakly excites
  16. L14:174 weakly suppresses
  17. L17:1887 weakly suppresses
  18. L21:560 weakly excites
  19. L14:1843 weakly excites
  20. L16:325 weakly suppresses
  21. L18:2974 weakly excites
  22. L18:1492 weakly excites
  23. L15:2251 weakly excites
  24. L13:313 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 -2.41 and -2.221 values between -2.04 and -1.854 values between -1.85 and -1.663 values between -1.66 and -1.484 values between -1.48 and -1.2917 values between -1.29 and -1.1124 values between -1.11 and -0.9250 values between -0.92 and -0.73115 values between -0.73 and -0.55187 values between -0.55 and -0.36413 values between -0.36 and -0.171509 values between -0.17 and 0.01943 values between 0.01 and 0.20339 values between 0.20 and 0.39140 values between 0.39 and 0.5764 values between 0.57 and 0.7638 values between 0.76 and 0.9524 values between 0.95 and 1.1314 values between 1.13 and 1.328 values between 1.32 and 1.516 values between 1.51 and 1.691 values between 1.69 and 1.881 values between 2.06 and 2.251 values between 2.62 and 2.811 values between 4.86 and 5.05offnegativepositiveoften

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. acting
  2. الخام
  3. bearing
  4. supremacy
  5. Each
  6. _)⏎
  7. ._⏎
  8. **⏎
  9. or
  10. evity

Suppresses

  1. posables
  2. asions
  3. ustom
  4. cket
  5. auses
  6. 远程
  7. urret
  8. conforms
  9. zeich
  10. życ