Neuron Explorer

Qwen3-0.6B

sorting and ordering words

A neuron in layer 4 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron lights up on words about putting things in order, such as 'order', 'sorted' and 'sort'. When it was turned up, the model tended to produce sequences and lists of numbers.

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

    . Three eighths is bigger than three ninths. Therefore three eighths is bigger than one third.⏎You arrange the data set in ascending order. You then find the observation such that a quarter of the observations are smaller than

  2. #2 peak

    .⏎Why is taxomony important to scientist?⏎Domain, Kingdom, Phylum, Class, Order, Family, Genus, Species⏎List the taxa in order from largest to smallest.⏎A heterotrophic eukaryote that absorbs nutrients

  3. #3 peak

    , and to manipulate it. ... To arrange atoms into an ordered nanotechnology landscape of precisely spaced steps and terraces. ... To determine the size

  4. #4 peak

    one a block then place them one after the other down the page. Consider a shop.⏎A shop might list the top selling product in each category with a view sorted by category. The marketing manager then asks for the categories in a different order

  5. #5 peak

    55 file extension⏎Recommended software programs are sorted by OS platform (Windows, Mac OS X etc.) and possible program actions⏎

  6. #6 peak

    rocks? Includes techniques. First of their formation. Fossils it contains compared to give an object. But with modification. Jeune femme - comparison between relative order in this radioactive. Determines the age of certain age of the house, as a

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 word "A" is being used in a box. So, what is this word? It is a box with the letter "A" (2010) The 1997-2010:

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

    purchase order for a large quantity of the product concerned. The basis of it is that goods

  2. #2 peak

    Board Class 10 Science Chapter 4 Environmental management questions answers. Our expert team have gone through this chapter and prepared Notes for the students who are facing difficulties in summarizing of

  3. #3 peak

    to teachers about how well students are performing in the concepts taught by them. If teachers find that most students are going wrong in a particular concept, they can arrange for remedial classes for such concepts and help students gain proficiency in the same. Extraction

  4. #4 peak

    spot and report errors or fraudulent transactions, just as you do with traditional banking.⏎2. Never give your Social Security number, credit or debit card numbers, personal identification numbers (PINs) or any other

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. L3:2028 suppresses
  2. L2:2856 suppresses
  3. L3:186 weakly suppresses
  4. L3:272 weakly excites
  5. L3:2369 weakly suppresses
  6. L3:1650 weakly suppresses
  7. L3:691 weakly excites
  8. L3:59 weakly excites
  9. L2:1036 weakly excites
  10. L1:2836 weakly suppresses
  11. L1:1674 weakly excites
  12. L2:1728 weakly suppresses
  13. L3:421 weakly suppresses
  14. L2:1892 weakly suppresses
  15. L3:122 weakly excites
  16. L3:2885 weakly excites
  17. L3:50 weakly excites
  18. L3:2941 weakly suppresses
  19. L0:3063 weakly suppresses
  20. L3:187 weakly excites
  21. L3:1696 weakly excites
  22. L3:1386 weakly excites
  23. L1:895 weakly excites
  24. L2:825 weakly excites

Feeds later layers

  1. L5:1160 excites
  2. L10:1207 excites
  3. L7:2463 excites
  4. L19:2312 excites
  5. L5:2140 excites
  6. L9:793 excites
  7. L26:2258 excites
  8. L18:2934 excites
  9. L7:2580 excites
  10. L24:2693 weakly excites
  11. L22:1605 weakly excites
  12. L24:2715 weakly excites
  13. L9:1365 weakly excites
  14. L7:1287 weakly excites
  15. L8:2642 weakly excites
  16. L23:1539 weakly suppresses
  17. L15:2566 weakly excites
  18. L7:1605 weakly excites
  19. L17:1246 weakly excites
  20. L25:1669 weakly excites
  21. L14:1824 weakly excites
  22. L7:892 weakly excites
  23. L13:98 weakly suppresses
  24. L5:1036 weakly suppresses

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 -0.64 and -0.591 values between -0.59 and -0.541 values between -0.48 and -0.439 values between -0.43 and -0.3812 values between -0.38 and -0.3233 values between -0.32 and -0.2770 values between -0.27 and -0.22161 values between -0.22 and -0.16301 values between -0.16 and -0.11552 values between -0.11 and -0.06853 values between -0.06 and -0.00866 values between -0.00 and 0.05476 values between 0.05 and 0.10280 values between 0.10 and 0.16171 values between 0.16 and 0.2166 values between 0.21 and 0.2630 values between 0.26 and 0.3212 values between 0.32 and 0.376 values between 0.37 and 0.422 values between 0.42 and 0.482 values between 0.48 and 0.531 values between 0.69 and 0.741 values between 0.90 and 0.961 values between 1.44 and 1.49offnegativepositiveoften

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. arranged
  2. getOrder
  3. ordered
  4. arrang
  5. Ascending
  6. .sorted
  7. essential
  8. ordered
  9. sắp
  10. Ordered

Suppresses

  1. )['
  2. ầm
  3. ound
  4. rossover
  5. antom
  6. unter
  7. ounds
  8. 불구하고
  9. �
  10. apters