Neuron Explorer

Qwen3-0.6B

'across' meaning spread over an area

A neuron in layer 1 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 lights up on the word 'across' when it means something is spread over a place, like 'across the country' or 'across Indiana'. Turning it up made the model's writing wander into general topics with no clear focus.

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

    become your organization's source for innovative, cost-saving, environmentally friendly proposals. There is a huge amount of information online, in local libraries and at thousands of universities across the country. The trouble is that much of the information, including cost studies,

  2. #2 peak

    “What Freedom Means to Me.” This annual event, sponsored by the Sertoma Club of Indiana, asks fifth- and sixth-grade students from participating schools all across Indiana

  3. #3 peak

    in COVID-19 deaths and cases is due to the spread of the Delta version in various nations.⏎“We continue to hear instances of hospitals reaching capacity from all across the world,” the DG stated.⏎“The Delta strain is spreading like wildfire over the

  4. #4 peak

    15,000 undergraduates, attracted more women and minorities and assisted nearly 50,000 science teachers and students at local schools across the country engaged in efforts to improve science curriculums.⏎Under the grants

  5. #5 peak

    for those who experience it. Unfortunately, despite efforts to raise awareness and implement preventative measures, instances of sexual harassment continue to occur in schools, colleges, and universities across the country.⏎In Pakistan, schools, colleges, and universities are considered to be

  6. #6 peak

    states and hundreds of communities across the country have already adopted strong bans on smoking, without any negative social or economic

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 number of a day's worth, and the number of a week's worth, and the number of a month's worth. I am sure that if a person is to work, then the person is to a month. So, 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

    , the first of Europe's Automated Transfer Vehicles, was launched from the ESA spaceport in French Guiana early March 9. It linked up with the space station

  2. #2 peak

    Society, around 12-20% women suffer from chronic pelvic pain all over the world. In certain rare cases, chronic pelvic pain is a strong indicator of more critical diseases like cancer. In

  3. #3 peak

    confusion and uncertainty, and can make it difficult for victims to seek help.⏎Another major issue is the lack of accountability for those who commit sexual harassment. In many cases, perpetrators are not

  4. #4 peak

    constriction and noisy breathing, while vocal cord surgery addresses similar problems with sound and articulation.⏎In many cases, surgeons operate on delicate structures such as minute blood vessels, which requires a level

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. L0:437 weakly suppresses
  2. L0:836 weakly suppresses
  3. L0:2300 weakly excites
  4. L0:432 weakly suppresses
  5. L0:2559 weakly suppresses
  6. L0:2412 weakly suppresses
  7. L0:2799 weakly suppresses
  8. L0:128 weakly excites
  9. L0:2196 weakly excites
  10. L0:1748 weakly excites
  11. L0:1415 weakly suppresses
  12. L0:2250 weakly excites
  13. L0:754 weakly suppresses
  14. L0:987 weakly suppresses
  15. L0:179 weakly suppresses
  16. L0:1657 weakly suppresses
  17. L0:1301 weakly suppresses
  18. L0:934 weakly suppresses
  19. L0:2145 weakly excites
  20. L0:475 weakly excites
  21. L0:2091 weakly suppresses
  22. L0:355 weakly suppresses
  23. L0:1224 weakly suppresses
  24. L0:2244 weakly suppresses

Feeds later layers

  1. L12:1811 weakly suppresses
  2. L15:520 weakly suppresses
  3. L7:1799 weakly excites
  4. L2:338 weakly excites
  5. L17:2504 weakly suppresses
  6. L8:1281 weakly suppresses uterus and womb anatomy words
  7. L18:943 weakly suppresses
  8. L25:180 weakly excites
  9. L16:662 weakly suppresses
  10. L23:28 weakly excites
  11. L14:2046 weakly excites scope words like all, many, suitable
  12. L11:91 weakly suppresses
  13. L22:1740 weakly excites
  14. L9:1949 weakly suppresses
  15. L7:839 weakly suppresses
  16. L13:644 weakly suppresses
  17. L16:317 weakly suppresses
  18. L24:2988 weakly excites
  19. L10:1668 weakly suppresses
  20. L3:402 weakly excites
  21. L20:1278 weakly suppresses
  22. L7:182 weakly suppresses
  23. L17:2756 weakly suppresses
  24. L8:3012 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.

1 values between -1.57 and -1.444 values between -0.90 and -0.764 values between -0.76 and -0.6316 values between -0.63 and -0.4934 values between -0.49 and -0.3677 values between -0.36 and -0.22297 values between -0.22 and -0.092269 values between -0.09 and 0.05850 values between 0.05 and 0.18195 values between 0.18 and 0.3269 values between 0.32 and 0.4532 values between 0.45 and 0.5816 values between 0.58 and 0.7215 values between 0.72 and 0.856 values between 0.85 and 0.995 values between 0.99 and 1.124 values between 1.12 and 1.264 values between 1.26 and 1.391 values between 1.39 and 1.533 values between 1.53 and 1.663 values between 2.07 and 2.201 values between 2.47 and 2.601 values between 2.87 and 3.011 values between 3.68 and 3.82offnegativepositiveoften

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. lider
  2. �数
  3. Lý
  4. ddl
  5. etri
  6. ере
  7. ky
  8. ери
  9. rax
  10. edly

Suppresses

  1. (EFFECT
  2. ><![
  3. 什么事情
  4. ContentValues
  5. SuppressWarnings
  6. CellValue
  7. .TabStop
  8. 什么事
  9. 再说
  10. apol