Neuron Explorer

Qwen3-0.6B

'of' after 'proportion' or 'percentage'

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 fires on the word 'of' when it follows 'proportion' or 'percentage', as in 'the proportion of teens'. Turning it up did not produce text clearly about proportions, so the output effect is hard to read.

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

    . But with economic development, the decrease in mortality rate becomes possible which ultimately increases the proportion of the younger population. The population pyramid becomes conic when we move upwards if other

  2. #2 peak

    of obesity increased 60 percent.⏎As a result, many schools have stopped selling pop or artificial juice to students.⏎Indeed, CDC data suggests that the proportion of teens who drink pop each day dropped from 29 percent in 20

  3. #3 peak

    for youth with mental disorders and the degree to which mental disorders of youth contribute to a community's overall delinquency. The second type of study examines the proportion of youth with mental disorders within juvenile justice facilities or programs. These studies provide information with

  4. #4 peak

    , to the ‘proportionate’ measure of poverty, we get a more meaningful insight into what happened to it. According to Planning Commission estimates, the proportion of the population below the poverty line in India was 44.5% in

  5. #5 peak

    of the female population, thereby indicating the sexes. Pyramids are of two types: One, based on absolute numbers of population and the second, on the proportion of population. The former shows the size and composition of population of a country. The

  6. #6 peak

    all.⏎Very few studies have used samples that make it possible to identify both the proportion of delinquent youth in a community who have mental disorders and the proportion of youth with mental disorders who<|endoftext|>COPD (chronic obstructive pulmonary disease

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 would like to think about this, I have been thinking about lately: I would like to think about this, I have been thinking about lately: I would like to think about this, I have been thinking about lately: I would like

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

    10.1177%2F1359104505053754<|endoftext|>From Sioux Falls to Rapid City, fire weather meteorologists are watching conditions closely. And until we receive widespread,

  2. #2 peak

    ّصَوُّف; صُوفِيّ ṣūfiyy / ṣūfī, مُتَصَوِّف mutaṣawwif), variously defined as

  3. #3 peak

    .⏎6. Adjust ties/straps as necessary to maintain cover and fit.⏎7. Untie straps or remove from behind the ears and pull the face covering away from your face touching only the ties/straps.⏎8. After

  4. #4 peak

    a high temperature, cooking food hot and fast. For thin cuts of meat and smaller items like kebabs, steaks, chops, burgers, and hot dogs, turn up the heat and cook them quickly. The fast pace means you need

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. L0:1081 weakly suppresses
  2. L0:1224 weakly suppresses
  3. L0:178 weakly suppresses
  4. L0:2224 weakly suppresses
  5. L0:1771 weakly suppresses
  6. L0:1301 weakly suppresses
  7. L0:866 weakly excites
  8. L0:2452 weakly suppresses
  9. L0:1075 weakly excites
  10. L0:1035 weakly excites
  11. L0:2019 weakly suppresses
  12. L0:1892 weakly excites
  13. L0:1915 weakly suppresses
  14. L0:2638 weakly excites
  15. L0:755 weakly suppresses
  16. L0:1750 weakly excites
  17. L0:1040 weakly excites
  18. L0:315 weakly excites
  19. L0:2370 weakly excites
  20. L0:1184 weakly suppresses
  21. L0:1706 weakly excites
  22. L0:1207 weakly suppresses
  23. L0:468 weakly excites
  24. L0:797 weakly suppresses

Feeds later layers

  1. L11:2337 excites
  2. L8:117 excites
  3. L8:300 weakly excites
  4. L10:2530 weakly excites
  5. L16:2448 weakly suppresses
  6. L3:959 weakly suppresses
  7. L17:2516 weakly excites
  8. L8:2960 weakly suppresses
  9. L11:2510 weakly suppresses
  10. L11:527 weakly excites
  11. L3:460 weakly suppresses
  12. L9:3017 weakly excites
  13. L10:2981 weakly excites
  14. L18:1375 weakly excites
  15. L13:1822 weakly suppresses
  16. L7:1146 weakly excites
  17. L16:1107 weakly suppresses
  18. L21:1958 weakly suppresses
  19. L4:1642 weakly suppresses
  20. L11:1795 weakly excites unclear: amounts, limits and 'right' answers
  21. L19:2761 weakly excites
  22. L14:1527 weakly suppresses
  23. L17:2177 weakly suppresses
  24. L22:1600 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 -1.30 and -1.201 values between -0.89 and -0.791 values between -0.79 and -0.691 values between -0.69 and -0.593 values between -0.59 and -0.483 values between -0.48 and -0.3818 values between -0.38 and -0.28179 values between -0.28 and -0.18743 values between -0.18 and -0.071896 values between -0.07 and 0.03818 values between 0.03 and 0.13160 values between 0.13 and 0.2336 values between 0.23 and 0.3414 values between 0.34 and 0.449 values between 0.44 and 0.545 values between 0.54 and 0.641 values between 0.64 and 0.754 values between 0.75 and 0.853 values between 0.85 and 0.952 values between 0.95 and 1.051 values between 1.05 and 1.161 values between 1.16 and 1.261 values between 1.36 and 1.463 values between 1.46 and 1.571 values between 1.67 and 1.771 values between 1.77 and 1.871 values between 2.08 and 2.181 values between 2.69 and 2.80offnegativepositiveoften

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. plants
  2. plates
  3. plant
  4. Issue
  5. opause
  6. imesteps
  7. weighed
  8. consideration
  9. .deleteById
  10. gatsby

Suppresses

  1. mvc
  2. rish
  3. MU
  4. )tableView
  5. nis
  6. apore
  7. apolis
  8. 白衣
  9. noc
  10. SqlDataAdapter