Neuron Explorer

Qwen3-0.6B

"factors" after "other" or "several"

A neuron in layer 2 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 word "factors" in phrases like "other factors that may increase your chance". It is one of the clearest patterns here, but turning it up did not produce text about factors, so the steering was not helpful.

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

    not causal. This means that the study did not determine a cause-and-effect relationship between early gluten ingestion and celiac, only that the two are connected.⏎Other factors influencing the study are that babies in Sweden tend to consume more foods that contain gluten

  2. #2 peak

    - In newborns, high levels of circulating estrogen may result in enlarged breast tissue and secretion of milk⏎Galactorrhea is more common in women. Other factors that may increase your chance of galactorr

  3. #3 peak

    for their unique ability to hold their breath underwater for extended periods, making them excellent divers. The duration of a turtle’s breath-holding capabilities varies depending on several factors, including the species of turtle, water temperature, and activity level. While some

  4. #4 peak

    -Mart store actually hinders a community’s ability to move families out of poverty.⏎After controlling for other factors that influence poverty rates, the researchers found that those U.S. counties in which

  5. #5 peak

    of women, and it depends on the stage of cancer at the time of detection.⏎But age isn’t the only risk factor for male breast cancer. Other possible factors include exposure to radiation, obesity, liver disease and Klinefelter Syndrome,

  6. #6 peak

    LOSE ENERGY?⏎Superior energy efficiency mostly depends on creating a better air seal. Wall insulation is a significant issue in home energy savings, but there are other factors that will lower home energy efficiency.⏎Doors and Windows⏎Water and moisture in

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 great person, I am a great man, I am a great man. I am a great man." So, I have been thinking about lately: "I am a great person, I am a great man, I

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

    ider, here to read about the award, and here to check out the project homepage.<|endoftext|>Since 1993, RAN’s Protect-an-Acre program (PAA

  2. #2 peak

    most of the sunshine. The school's large roof generates more solar power than the school consumes, especially in the summer months and during weekends. The NPO APERe (Association pour la Promotion des Energies Renouvables) and Brussels grid

  3. #3 peak

    rainforests and that frontline communities organizing against the extraction and burning of dirty fossil fuels deserve the strongest support we can offer. RAN established the Protect-an-Acre program to protect the world’s forests and the rights of their inhabitants by providing financial

  4. #4 peak

    communities suffer disproportionate impacts to their health, livelihood and culture from extractive industry mega-projects and the effects of global climate change. That’s why Protect-an-Acre provides small grants to community-based organizations, Indigenous federations and small NGOs that are

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. L0:984 weakly excites
  2. L0:924 weakly suppresses
  3. L1:418 weakly excites "of" after "components" or "causes"
  4. L0:1936 weakly suppresses
  5. L1:340 weakly excites
  6. L0:2173 weakly excites
  7. L0:974 weakly excites
  8. L0:228 weakly suppresses
  9. L1:1739 weakly excites
  10. L0:743 weakly excites
  11. L1:1986 weakly excites
  12. L1:2819 weakly suppresses
  13. L0:1353 weakly suppresses
  14. L1:2342 weakly excites
  15. L0:2853 weakly excites
  16. L1:801 weakly excites
  17. L1:2023 weakly suppresses
  18. L1:1798 weakly excites
  19. L1:776 weakly excites
  20. L0:1049 weakly excites
  21. L0:773 weakly excites
  22. L1:2856 weakly suppresses
  23. L1:1957 weakly excites
  24. L1:83 weakly suppresses

Feeds later layers

  1. L3:2036 weakly suppresses
  2. L9:2943 weakly excites
  3. L18:801 weakly excites
  4. L23:1575 weakly suppresses
  5. L5:495 weakly suppresses
  6. L6:2451 weakly excites
  7. L9:196 weakly suppresses
  8. L8:1999 weakly excites
  9. L11:693 weakly suppresses
  10. L5:1474 weakly suppresses
  11. L9:477 weakly suppresses
  12. L4:1719 weakly suppresses
  13. L14:1035 weakly excites
  14. L8:318 weakly suppresses
  15. L3:1471 weakly suppresses
  16. L21:891 weakly suppresses
  17. L9:2772 weakly excites
  18. L25:2817 weakly suppresses
  19. L11:1852 weakly suppresses the digit 1 in numbers
  20. L26:875 weakly suppresses
  21. L22:191 weakly excites
  22. L13:2752 weakly suppresses
  23. L10:141 weakly excites
  24. L8:1635 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 -1.21 and -1.111 values between -1.00 and -0.891 values between -0.79 and -0.682 values between -0.68 and -0.584 values between -0.58 and -0.476 values between -0.47 and -0.3725 values between -0.37 and -0.26132 values between -0.26 and -0.16577 values between -0.16 and -0.052064 values between -0.05 and 0.06809 values between 0.06 and 0.16220 values between 0.16 and 0.2740 values between 0.27 and 0.376 values between 0.37 and 0.485 values between 0.48 and 0.582 values between 0.58 and 0.692 values between 0.69 and 0.793 values between 0.79 and 0.902 values between 0.90 and 1.011 values between 1.01 and 1.112 values between 1.11 and 1.221 values between 1.53 and 1.641 values between 2.17 and 2.271 values between 2.90 and 3.01offnegativepositiveoften

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. unik
  2. rio
  3. {@
  4. attrib
  5. ḿ
  6. 禀
  7. ач
  8. airo
  9. 重要因素
  10. knob

Suppresses

  1. gọn
  2. あたり
  3. しさ
  4. cruelty
  5. juste
  6. зе
  7. 練
  8. 暢
  9. -lasting
  10. /Gate