Neuron Explorer

Qwen3-0.6B

the word 'whole' meaning entire

A neuron in layer 23 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 responds to the word 'whole' in phrases like 'the whole family' or 'a whole year', meaning 'entire'. When it was turned down, the model got stuck repeating 'whole'.

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

    support<|endoftext|>Extra 10% off sitewide + Free shipping over $150 with code: STOCKUP Details⏎Happy Homeschool Month! The whole month of January we will have NEW DEALS every week for you to stock up

  2. #2 peak

    has in your child’s health and well-being. But their help can extend beyond regular check-ups and immunizations. Pediatricians and their support staff can help your whole family lead a healthier lifestyle. Let’s begin with understanding the role of your child

  3. #3 peak

    embarrassed. Don’t be concerned. Your pediatrician has heard it all before and can give professional advice while providing you with peace of mind.⏎- Advice for the whole family: Though a pediatrician’s focus is children, they are

  4. #4 peak

    is⏎more than enough energy to satisfy the global energy needs for a⏎whole year. However, the technology produces less than one tenth of⏎one percent of

  5. #5 peak

    over-talkative.⏎There are many things that can be involved when children have problems with behavior, feelings or relationships. They can be related to the child or the whole family and can include:⏎- family history of mental health problems⏎- serious illness

  6. #6 peak

    . Attributable to this, the forest officials can not only restrict tribal and forest dwellers from accessing forest lands and produce but can also punish the whole community for the act of an individual.⏎Into the bargain, the government has gone

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 whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole whole

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

    pace.⏎Also See | The Recession Alphabet (Graphics)⏎The deep 1973-75 recession saw only a mild decline in economic activity for almost the first year of recession, but was then followed by a plunge and a sharp

  2. #2 peak

    be redefined. Instead of searching the optimal solution for the individual machines (engine, turbocharger, power turbine, boiler and steam turbine), a combined optimum for the entire process

  3. #3 peak

    3 states studied but increased for females in 7 states and males in 5 states. Life expectancy improved but remains lowest for non-Hispanic Black people for almost every state.⏎An accompanying editorial, authors from Washington University and the University

  4. #4 peak

    .....Of interest to serious advanced students of cognitive development and epistemologists."<|endoftext|>Whangamarino Wetland is a lowland mosaic of wetland types covering almost 7,000 ha. It is the second

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. L15:1819 suppresses
  2. L20:2457 excites
  3. L22:2864 suppresses
  4. L20:1862 suppresses
  5. L18:2955 excites
  6. L20:1219 suppresses
  7. L19:2964 suppresses
  8. L15:1535 suppresses
  9. L20:1945 suppresses
  10. L19:1904 suppresses
  11. L6:2825 excites
  12. L22:1015 excites
  13. L5:3056 suppresses
  14. L22:477 suppresses
  15. L20:3006 suppresses
  16. L7:2879 excites
  17. L18:2375 excites
  18. L17:2395 suppresses
  19. L19:1661 excites
  20. L17:828 suppresses
  21. L21:2170 weakly suppresses
  22. L19:3065 weakly excites 'for' after a span of time
  23. L13:2164 weakly excites
  24. L17:2514 weakly excites

Feeds later layers

  1. L26:684 strongly suppresses
  2. L25:2606 strongly suppresses
  3. L24:1375 strongly suppresses
  4. L27:1427 suppresses
  5. L27:1812 suppresses
  6. L27:127 excites
  7. L27:399 weakly suppresses
  8. L26:507 weakly suppresses
  9. L27:823 weakly suppresses
  10. L27:1840 weakly suppresses
  11. L27:1752 weakly suppresses
  12. L24:2445 weakly excites
  13. L25:105 weakly excites
  14. L24:2340 weakly suppresses
  15. L25:924 weakly excites
  16. L25:530 weakly suppresses
  17. L25:1185 weakly suppresses
  18. L26:913 weakly suppresses
  19. L26:1600 weakly suppresses
  20. L24:2353 weakly suppresses
  21. L24:1666 weakly excites
  22. L25:1947 weakly suppresses
  23. L24:1373 weakly suppresses 'all' in 'all over the world'
  24. L24:1778 weakly suppresses 'a' before a time span

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.

3 values between -14.65 and -13.991 values between -12.68 and -12.032 values between -9.41 and -8.754 values between -6.79 and -6.135 values between -6.13 and -5.485 values between -5.48 and -4.823 values between -4.82 and -4.161 values between -4.16 and -3.517 values between -3.51 and -2.8516 values between -2.85 and -2.2028 values between -2.20 and -1.5464 values between -1.54 and -0.89421 values between -0.89 and -0.232353 values between -0.23 and 0.42886 values between 0.42 and 1.0869 values between 1.08 and 1.7312 values between 1.73 and 2.399 values between 2.39 and 3.045 values between 3.04 and 3.702 values between 3.70 and 4.354 values between 4.35 and 5.014 values between 5.01 and 5.672 values between 6.98 and 7.631 values between 7.63 and 8.291 values between 10.91 and 11.56offnegativepositiveoften

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. 低估
  2. ropriate
  3. efault
  4. avern
  5. 之一
  6. ridge
  7. NF
  8. acons
  9. ordination
  10. bling

Suppresses

  1. whole
  2. whole
  3. entire
  4. 整个
  5. 整
  6. Whole
  7. Whole
  8. Entire
  9. 全
  10. 整個