Neuron Explorer

Qwen3-0.6B

comma or bracket after a name in a byline

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

Fires on the comma or closing bracket that comes right after a person's name at the end of a quote or social media post, as in a credit line. Turning it up gave text that kept shifting between topics, so the steering does not show this pattern clearly.

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

    based on real data, with something familiar for scale pic.twitter.com/OdQA07qhBb⏎— Adrian Luckman (@adrian_luckman) July 12, 2017⏎You may imagine that the

  2. #2 peak

    females have crept into their privileges, but I can’t help the poor fellows, they will have to get used to it.Henriette Hirschfeld, circa 1886.⏎Henriette Hirschfeld (18

  3. #3 peak

    NGA).” My question is this: When did Montana begin to be represented by associations and councils rather than our elected representatives in the legislature? Denise Juneau, our state superintendent, went to the Montana Board of Public Education

  4. #4 peak

    surely be conducted in the coming years but at this point, you’re safe to use all the free night and weekend minutes that you want.⏎Adam Brandeberry, Med IV (OSU COM)⏎John A. Vaughn, MD (OS

  5. #5 peak

    the Moxy people. I did a post and review on that article. https://t.co/D9NIorgXuc— Bruce Rogers (@bjrmd) August 30, 2021⏎PACING: Dist

  6. #6 peak

    view from our vehicles. This photographic guide can help you discover and learn about the plants inhabitating our northern prairies and plains." David J. Ode, Botanist/Ecologist, South

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 first paragraph in this sentence is the first sentence of the book, and the first sentence of the first sentence is the first sentence of the first sentence is the first sentence of the first sentence is the first sentence of the first sentence is the

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

    O’Neil was declared a conservation area.⏎Neil burn Drive was created.⏎Canmore place was erected.<|endoftext|>ERIC Number: ED379961⏎Record Type: RIE⏎Publication Date: 19

  2. #2 peak

    case of further problems read the IDEAS help page. Note that these files are not on the IDEAS site. Please be patient as the files may be large.⏎As the access to this document is restricted, you may want to look for

  3. #3 peak

    ⏎- Other factors (e.g. extremes in temperatures, sunlight, tight-fitting clothing, etc.)<|endoftext|>ERIC Number: ED348648⏎Record Type: Non-Journal⏎Publication Date: 1992-Apr

  4. #4 peak

    (8), 1595-1600. http://dx.doi.org/10.1016/j.jbusres.2013.10.001⏎BERISHA-N

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. L10:36 excites
  2. L11:2366 excites
  3. L6:1053 weakly excites
  4. L1:1850 weakly excites percent-encoded url characters
  5. L7:803 weakly suppresses
  6. L3:1123 weakly excites
  7. L3:972 weakly excites
  8. L11:2434 weakly excites
  9. L4:485 weakly excites
  10. L10:1006 weakly excites
  11. L3:669 weakly suppresses
  12. L5:262 weakly suppresses
  13. L11:161 weakly excites
  14. L3:874 weakly excites
  15. L7:1027 weakly excites
  16. L4:2166 weakly excites
  17. L2:509 weakly suppresses
  18. L2:837 weakly suppresses
  19. L11:157 weakly excites
  20. L7:2417 weakly excites
  21. L8:254 weakly excites
  22. L11:39 weakly excites
  23. L3:2857 weakly suppresses the word 'fiber' or 'fibre'
  24. L7:1667 weakly suppresses

Feeds later layers

  1. L13:52 excites
  2. L13:1305 weakly suppresses
  3. L16:2282 weakly suppresses
  4. L13:1650 weakly suppresses
  5. L13:233 weakly excites
  6. L13:915 weakly excites
  7. L13:1101 weakly excites
  8. L20:2392 weakly suppresses
  9. L17:225 weakly excites
  10. L18:2573 weakly excites
  11. L15:2231 weakly excites
  12. L16:378 weakly suppresses
  13. L13:190 weakly suppresses
  14. L17:592 weakly suppresses
  15. L15:590 weakly suppresses
  16. L17:2885 weakly suppresses
  17. L13:2672 weakly suppresses
  18. L26:884 weakly suppresses
  19. L17:308 weakly excites
  20. L15:1299 weakly excites
  21. L16:322 weakly excites
  22. L13:2690 weakly suppresses
  23. L20:2170 weakly suppresses
  24. L13:773 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 -0.99 and -0.941 values between -0.74 and -0.692 values between -0.64 and -0.593 values between -0.59 and -0.544 values between -0.54 and -0.496 values between -0.49 and -0.4417 values between -0.44 and -0.3932 values between -0.39 and -0.3473 values between -0.34 and -0.29118 values between -0.29 and -0.24224 values between -0.24 and -0.19321 values between -0.19 and -0.14453 values between -0.14 and -0.09492 values between -0.09 and -0.04674 values between -0.04 and 0.01548 values between 0.01 and 0.06355 values between 0.06 and 0.11247 values between 0.11 and 0.16151 values between 0.16 and 0.21102 values between 0.21 and 0.2640 values between 0.26 and 0.3123 values between 0.31 and 0.368 values between 0.36 and 0.415 values between 0.41 and 0.462 values between 0.46 and 0.511 values between 0.61 and 0.663 values between 0.80 and 0.851 values between 0.85 and 0.901 values between 0.95 and 1.00offnegativepositiveoften

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. ites
  2. ych
  3. ��态
  4. ilter
  5. s
  6. amma
  7. ใจ
  8. time
  9. ifth
  10. ajor

Suppresses

  1. iday
  2. CATEGORY
  3. Crow
  4. =zeros
  5. Crow
  6. nov
  7. .lua
  8. �回
  9. �
  10. 未成