Neuron Explorer

Qwen3-0.6B

"self" before esteem or image

A neuron in layer 24 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 "self" when it comes before "esteem", "image" or "worth", as in self-esteem. When turned up, the model kept producing the word "image".

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

    for the hard work you ask them to do when you train and compete, an unhealthy focus on these elements of yourself can have negative impacts on your mental health and self esteem.⏎Some actions and thoughts to consider;⏎Don't compare yourself to others,

  2. #2 peak

    will is crushed by suppressive parents, they could end up depressed, anxious or ineffective in their life. A thwarted free will can significantly impact the child’s self-worth.⏎During earlier stages of human society, compliance was a positive and necessary

  3. #3 peak

    give us something to look forward to, provide a social outlet, allay worries about health and fitness, increase energy levels, and improve self-esteem. All these factors can lessen the severity of anxiety and depression. Exercise also

  4. #4 peak

    UNICEF show that participation in athletics reaps countless long-term benefits for girls. For example, playing sports has immense psychological benefits for girls, such as improved self image, increased self confidence, positive feelings and knowledge about one’s body, and

  5. #5 peak

    of suicide. Even years after the bullying has stopped, researchers have found that adults who were bullied as teens have higher levels of depression and poorer self-esteem than other adults.⏎A recent study that<|endoftext|>|Cherokee County, Texas

  6. #6 peak

    1,148% growth rate.⏎These children endure stigmatization and/or bullying on a daily basis. Many are emotionally shattered; most have a poor self-image and ALL are vulnerable. In addition, many of the

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: when the call bously called the test, the test had a call from the call from the call, which called the call back, then the call from the call from the call, which called back the call, then the call from 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

    CE) saw the rise of Zen Buddhism in Japan, which greatly influenced Japanese art. Zen-inspired ink paintings and tea ceremony ceramics became popular during this time.⏎The Edo period (1603 CE to 1868 CE)

  2. #2 peak

    aren’t supposed to be greedy for gain (1 Tim. 3:8). That may suggest they were responsible for the finances of the church

  3. #3 peak

    the young pastor (1 Timothy 1:18). God continues to use children today in their schools, homes

  4. #4 peak

    a genre of Japanese art that flourished during the Edo period (1603-1868). It is a

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. L4:1924 weakly suppresses
  2. L5:2636 weakly excites
  3. L19:1385 weakly excites
  4. L20:2513 weakly suppresses
  5. L21:1539 weakly excites
  6. L17:2391 weakly excites
  7. L18:1376 weakly suppresses the word 'case' in set phrases
  8. L23:2629 weakly suppresses
  9. L19:929 weakly excites
  10. L19:130 weakly suppresses
  11. L23:2090 weakly excites
  12. L13:1731 weakly suppresses
  13. L18:336 weakly suppresses
  14. L10:411 weakly suppresses
  15. L8:920 weakly excites
  16. L23:649 weakly excites
  17. L22:2555 weakly suppresses
  18. L6:1206 weakly excites
  19. L5:218 weakly suppresses
  20. L7:406 weakly suppresses
  21. L11:816 weakly excites
  22. L9:1024 weakly excites
  23. L21:381 weakly excites zeros inside long number codes
  24. L4:101 weakly excites

Feeds later layers

  1. L27:1316 strongly excites
  2. L25:1584 excites art and artist names in lists
  3. L26:558 weakly excites
  4. L25:147 weakly suppresses
  5. L25:1895 weakly suppresses first digit of a year like 2020
  6. L27:902 weakly excites
  7. L27:768 weakly excites
  8. L26:1218 weakly suppresses
  9. L25:322 weakly suppresses
  10. L27:1235 weakly excites
  11. L26:1562 weakly suppresses
  12. L25:342 weakly excites
  13. L27:2371 weakly excites
  14. L25:995 weakly suppresses
  15. L25:845 weakly suppresses
  16. L27:1389 weakly suppresses
  17. L26:834 weakly excites
  18. L27:352 weakly excites
  19. L27:2066 weakly excites
  20. L27:2150 weakly excites
  21. L25:224 weakly suppresses
  22. L27:167 weakly suppresses
  23. L26:1384 weakly suppresses
  24. L27:1525 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 -10.84 and -9.408 values between -9.40 and -7.954 values between -7.95 and -6.507 values between -6.50 and -5.058 values between -5.05 and -3.6046 values between -3.60 and -2.15191 values between -2.15 and -0.703360 values between -0.70 and 0.75215 values between 0.75 and 2.2036 values between 2.20 and 3.6512 values between 3.65 and 5.106 values between 5.10 and 6.553 values between 6.55 and 8.001 values between 8.00 and 9.452 values between 9.45 and 10.902 values between 10.90 and 12.341 values between 15.24 and 16.691 values between 18.14 and 19.591 values between 19.59 and 21.041 values between 21.04 and 22.491 values between 22.49 and 23.941 values between 45.68 and 47.13offnegativepositiveoften

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. image
  2. Image
  3. images
  4. image
  5. 形象
  6. (image
  7. 图像
  8. →image
  9. -image
  10. _image

Suppresses

  1. thesize
  2. boil
  3. retty
  4. ัน
  5. 줄
  6. Dot
  7. adal
  8. Middleware
  9. 鹩
  10. assen