Neuron Explorer

Qwen3-0.6B

unclear: mixed mid-sentence words

A neuron in layer 8 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on a grab bag of ordinary words in the middle of sentences, with no clear shared meaning. Turning it up made the model's text drift into unrelated topics and repetitive loops, so it doesn't seem to control one clear behaviour.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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

    511 Fourth Avenue Southeast in 1858. Woodbury Fisk, also a native of New Hampshire, came to St. Anthony in 1856 and became a principle in the flour milling firm of Pillsbury

  2. #2 peak

    , Jane Hurley Wilson ’64 and Michael G. Wilson, and the Getty Foundation.⏎Daniela Gonzalez-Pruitt, University of Toronto ’17 and Josephine Ren, SCR ’19⏎Stories Behind the Scenes⏎In his

  3. #3 peak

    one of these, and imagine dividing it into 10 pieces. Then take one of those pieces, and divide it into 10 more pieces. Now take one of those pieces, if you can find one, and divide it, once

  4. #4 peak

    sugar levels. Ginger, another key spice in pumpkin pie, is well-known for its benefits as a digestive aid and is particularly helpful for nausea. Cloves can benefit digestion and the immune system, and nutmeg can

  5. #5 peak

    can be compared with the expected number of answer matches. Example: John (62 correct out of 100) and Barbara (67 correct out of 100) answer 89 questions exactly the same way. Scan

  6. #6 peak

    15||ARAB 200 AND ARAB 300⏎|FALL 2015||ARAB 100 AND ARAB 210|⏎|SPRING 2016||

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 am a person who is not a man of the world, but I am not a man of the world." So I have to be in the room with the other students, and I am in the room with the other students,

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

    in Orange County, California have been professionally diagnosed with autism, Down syndrome or some form of developmental disability.⏎According to the U.S. Census Bureau, more than 1 in 150 children are diagnosed as having an Autism Spectrum Disorder.

  2. #2 peak

    , either as a victim, a perpetrator, or both. The American Academy of Child & Adolescent Psychiatry believes the problem is even more widespread, reporting that as many as 50 percent of all young people are bullied at least once during their school

  3. #3 peak

    Scotland, and granddaughter of Kenneth III .. At the instigation of her father Gruoch was married to Gillacomagnan Earl of Moray. The son of this union, Lulach, followed his stepfather to the throne of Scotland

  4. #4 peak

    19% with prostate cancer; and 10% with colorectal cancer.⏎Those three cancers accounted for slightly more than half of all cancer diagnoses.⏎Nearly 60% of survivors were 65 or older, the report found.

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. L7:667 suppresses
  2. L7:97 suppresses
  3. L6:135 excites
  4. L5:82 excites
  5. L7:482 suppresses
  6. L7:542 excites
  7. L7:88 excites
  8. L6:1180 excites
  9. L5:29 suppresses
  10. L7:333 suppresses
  11. L7:75 excites
  12. L6:210 suppresses
  13. L6:1554 suppresses
  14. L7:872 excites
  15. L2:178 excites
  16. L5:2654 suppresses
  17. L6:484 suppresses
  18. L5:154 suppresses
  19. L6:266 excites
  20. L5:345 excites
  21. L6:51 excites
  22. L6:1064 excites
  23. L7:33 excites
  24. L5:295 suppresses

Feeds later layers

  1. L11:1100 suppresses
  2. L9:78 suppresses
  3. L13:321 excites
  4. L9:759 excites
  5. L12:1520 suppresses
  6. L11:1037 suppresses
  7. L11:1907 suppresses
  8. L10:385 suppresses
  9. L14:2227 suppresses
  10. L12:1029 suppresses
  11. L13:582 suppresses
  12. L9:1574 suppresses
  13. L9:700 excites
  14. L13:1109 suppresses
  15. L9:2419 suppresses
  16. L12:557 excites
  17. L12:723 suppresses
  18. L11:1282 suppresses
  19. L10:1402 suppresses
  20. L10:415 suppresses
  21. L9:1010 suppresses
  22. L17:600 suppresses
  23. L9:2055 suppresses
  24. L12:590 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 -1.98 and -1.891 values between -1.60 and -1.501 values between -1.50 and -1.411 values between -1.41 and -1.311 values between -1.31 and -1.212 values between -1.21 and -1.123 values between -1.12 and -1.025 values between -1.02 and -0.934 values between -0.93 and -0.8310 values between -0.83 and -0.7315 values between -0.73 and -0.6429 values between -0.64 and -0.5436 values between -0.54 and -0.4557 values between -0.45 and -0.35128 values between -0.35 and -0.25252 values between -0.25 and -0.16606 values between -0.16 and -0.061808 values between -0.06 and 0.03609 values between 0.03 and 0.13164 values between 0.13 and 0.2374 values between 0.23 and 0.3244 values between 0.32 and 0.4217 values between 0.42 and 0.519 values between 0.51 and 0.6111 values between 0.61 and 0.717 values between 0.71 and 0.806 values between 0.80 and 0.902 values between 0.90 and 0.992 values between 0.99 and 1.091 values between 1.09 and 1.181 values between 1.28 and 1.381 values between 1.76 and 1.86offnegativepositiveoften

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. scape
  2. otta
  3. озд
  4. elige
  5. GetInstance
  6. obb
  7. ospel
  8. 王爷
  9. disproportion
  10. Scalars

Suppresses

  1. ecast
  2. haar
  3.  ̄
  4. do
  5. 豆腐
  6. uria
  7. .des
  8. 绿色通道
  9. -door
  10. -Un