Neuron Explorer

Qwen3-0.6B

'all' or 'every' before a group

A neuron in layer 7 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 reacts to the word "all" or "every" when it comes before a group of things, like "all types of oaks" or "every continent". Its weaker negative side fires on words like "otherwise" and "ever". Turning it up gave text that drifted off topic and did not show the pattern clearly.

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

    . For example, Maples are plagued by Maple Spider Mites and Honeylocust by Honeylocust Spider Mites.⏎Oak Spider Mites in particular attack all types of Oaks; however, pin oaks and willow oaks tend to

  2. #2 peak

    as a natural meat tenderizer. In addition, bromelain may also induce a prickly sensation in the mouth when consumed.⏎Although the enzyme is present in all parts of the pineapple plant, it's most concentrated in the stem, the hard

  3. #3 peak

    going through changes into your 20s, you generally gain more maturity when you are 18 or<|endoftext|>Menopause is a natural process that occurs in all women when they are older. This stage may cause several unpleasant effects that make your

  4. #4 peak

    involving both prescribed and illicit opioids are rising sharply. They’ve nearly doubled since 2009, and have infiltrated all genders, demographics and geographic areas, according to Centers for Disease Control and Prevention

  5. #5 peak

    problems are serious but others are not.⏎Age-related changes in memory⏎Forgetfulness can be a normal part of aging. As people get older, changes occur in all parts of the body, including the brain. As a result, some people may

  6. #6 peak

    implementing the tsujigahana, a 16th-century technique that produces the most obscure and jaw-dropping masterpieces. He used this technique in all of his kimono creations. When the Kubota Museum was almost bankrupt, 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: the role of a teacher in the field of education, the development of new technologies, and the impact of social movements. For example, I have seen that a teacher's work might be seen as controversial, leading to debates about the legitimacy of

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

    CARE.)⏎In the dark all cats are gray. All persons are undistinguished till they have made a name.⏎Not room to swing a cat. Swinging

  2. #2 peak

    ont Abbey in Normandy and they established a small priory there. It only ever housed two monks however and seems to have been a simple manor house plus chapel

  3. #3 peak

    the⏎major arc AB,⏎where you go the long way around the bottom of the circle. Unless stated otherwise, it always means the⏎minor arc - the shortest of the two.⏎If

  4. #4 peak

    . The expression “Vale un potosi!” (it costs a potosi!), from Miguel de Cervantes, had finished with the reputation of the queen of the American cities. It represents the symbol of the New World and the Golden Age

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. L6:1461 suppresses
  2. L5:2789 suppresses
  3. L5:2179 suppresses "the" after "one of"
  4. L6:426 suppresses
  5. L4:262 excites
  6. L4:2143 suppresses "in some" opening a qualifier
  7. L6:2021 excites
  8. L3:2187 suppresses
  9. L6:1363 excites
  10. L5:225 weakly excites
  11. L6:380 weakly excites
  12. L6:1655 weakly excites "a" or "an" after "such" or launch verbs
  13. L6:2112 weakly excites
  14. L4:661 weakly excites
  15. L5:2448 weakly suppresses
  16. L4:141 weakly excites
  17. L5:2519 weakly suppresses
  18. L6:2914 weakly suppresses
  19. L4:272 weakly suppresses
  20. L4:1995 weakly suppresses
  21. L6:2867 weakly excites
  22. L5:1830 weakly excites
  23. L3:1700 weakly excites
  24. L5:2545 weakly suppresses the zero in round numbers like 100

Feeds later layers

  1. L9:620 suppresses
  2. L12:1155 suppresses
  3. L16:1273 suppresses
  4. L12:942 excites
  5. L11:1517 excites
  6. L15:2651 excites
  7. L11:522 suppresses
  8. L15:1925 excites
  9. L17:2080 suppresses
  10. L12:491 excites
  11. L9:352 weakly suppresses
  12. L13:790 weakly suppresses
  13. L20:2810 weakly excites
  14. L13:51 weakly suppresses
  15. L9:918 weakly excites
  16. L9:543 weakly excites
  17. L14:380 weakly suppresses
  18. L9:781 weakly suppresses
  19. L9:334 weakly excites
  20. L10:633 weakly suppresses
  21. L11:1674 weakly suppresses
  22. L11:267 weakly excites
  23. L13:2005 weakly excites
  24. L8:2310 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 -2.39 and -2.231 values between -2.08 and -1.933 values between -1.33 and -1.184 values between -1.18 and -1.038 values between -1.03 and -0.8712 values between -0.87 and -0.727 values between -0.72 and -0.5759 values between -0.57 and -0.42220 values between -0.42 and -0.27630 values between -0.27 and -0.121973 values between -0.12 and 0.03648 values between 0.03 and 0.18166 values between 0.18 and 0.3468 values between 0.34 and 0.4935 values between 0.49 and 0.6421 values between 0.64 and 0.7919 values between 0.79 and 0.945 values between 0.94 and 1.095 values between 1.09 and 1.245 values between 1.24 and 1.395 values between 1.39 and 1.544 values between 1.54 and 1.695 values between 1.69 and 1.852 values between 2.60 and 2.751 values between 2.90 and 3.051 values between 3.51 and 3.66offnegativepositiveoften

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. always
  2. 百分百
  3. тельные
  4. 迪
  5. correct
  6. LEGRO
  7. Prim
  8. ippers
  9. diaper
  10. always

Suppresses

  1. ämpfe
  2. ]</
  3. 尤为重要
  4. 硖
  5. ,</
  6. ircles
  7. 他还
  8. ueva
  9. %</
  10. nữa