Neuron Explorer

Qwen3-0.6B

digits inside numbers

A neuron in layer 10 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 lights up on the digits inside numbers, especially the zeros and last digits that finish a figure like 10,000 or $10. When it was turned up, the written text drifted into unrelated topics, so the steering did not show a clear effect.

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

    of the year using altars of various geometric configurations. India is largely credited with inventing the idea of zero and for developing place notation (ones, tens, hundreds, etc.), the decimal system, and the numeral system used today.⏎Already established

  2. #2 peak

    1 seconds). As a comparison, today’s hard drives take a few nanoseconds to record 1 bit of data. A nanosecond is 1000 picoseconds, making it very clear how much faster the heat pulse solution will

  3. #3 peak

    names organisms based on studies of their different characteristics. Helps to study the relationships between living and extinct species.⏎Why is taxomony important to scientist?⏎Domain, Kingdom, Phylum, Class, Order, Family, Genus, Species⏎List

  4. #4 peak

    55, a re-evaluation took place and a new Yuan Renminbi was introduced at a rate of 1 new Yuan to 10,000 old Yuan.⏎The Renminbi in Foreign Exchange⏎During the command economy

  5. #5 peak

    that may be linked to cancer (nonmuscle myosin II). In addition, inherited familial cardiomyopathies affecting 1 out of 500 people result from missense mutations in various cardiac muscle proteins, with mutations in beta

  6. #6 peak

    Here the children are asked to look at the grid of numbers and identify the missing numbers and then fill in the missing numbers to count from 1 to 25. In these missing number worksheets some of the numbers are already filled in to help

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 one that is... if you look at it. The first thing that comes to mind is that, for example, and then, and then, and so on. The problem, the first thing, and the second, and the third

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

    and don’t cut back on food and drink, we need to do more.⏎The American College of Sports Medicine agrees adults should get at least 150 minutes of exercise a week, though it explains this might be 20 to

  2. #2 peak

    at the time of a maltreatment investigation associated with the likelihood of remaining in foster care? NSCAW is a national longitudinal study of 5,501 children, aged 14 years or younger, who had contact with the child

  3. #3 peak

    July 10, the fire had damaged or destroyed nearly 130 buildings, injured over 20 people, and cut the lives of 19 Granite Mountain Hotshots woefully short.⏎Sparked by Lightning, Fueled

  4. #4 peak

    against similar survey data from 2009 to 2012.⏎Study participants included 570 students in 8th grade (49.6% girls; mean age, 13.5 years),

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. L8:41 suppresses
  2. L9:657 suppresses
  3. L7:895 suppresses
  4. L0:208 suppresses the digit zero inside codes or numbers
  5. L1:1993 weakly excites
  6. L2:241 weakly excites
  7. L5:364 weakly suppresses
  8. L3:2824 weakly suppresses
  9. L4:1926 weakly excites
  10. L9:2379 weakly excites
  11. L9:3037 weakly excites
  12. L4:1343 weakly excites
  13. L6:567 weakly suppresses technical terms mid-phrase, mixed topics
  14. L4:394 weakly excites
  15. L8:1706 weakly suppresses
  16. L3:279 weakly excites
  17. L6:2354 weakly excites
  18. L3:230 weakly excites
  19. L1:2776 weakly suppresses
  20. L9:1926 weakly suppresses
  21. L7:1753 weakly suppresses
  22. L0:1788 weakly excites
  23. L5:424 weakly suppresses
  24. L4:667 weakly excites

Feeds later layers

  1. L13:2568 excites
  2. L12:946 weakly excites
  3. L15:2117 weakly excites
  4. L17:1522 weakly suppresses
  5. L12:1194 weakly excites
  6. L15:2768 weakly suppresses
  7. L18:1125 weakly excites
  8. L11:2080 weakly suppresses
  9. L11:1106 weakly excites
  10. L17:1795 weakly suppresses 'and' before a second verb phrase
  11. L21:734 weakly suppresses
  12. L23:615 weakly excites
  13. L27:1131 weakly excites
  14. L17:1580 weakly excites
  15. L12:2550 weakly suppresses
  16. L13:1019 weakly excites
  17. L24:3048 weakly suppresses
  18. L14:263 weakly excites
  19. L15:1066 weakly suppresses
  20. L17:3065 weakly excites
  21. L21:1244 weakly suppresses
  22. L13:2869 weakly suppresses
  23. L23:2121 weakly excites
  24. L12:1272 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 -2.26 and -2.141 values between -2.14 and -2.022 values between -2.02 and -1.901 values between -1.90 and -1.774 values between -1.53 and -1.411 values between -1.41 and -1.281 values between -1.28 and -1.167 values between -1.16 and -1.048 values between -1.04 and -0.9212 values between -0.92 and -0.799 values between -0.79 and -0.6722 values between -0.67 and -0.5524 values between -0.55 and -0.4352 values between -0.43 and -0.31135 values between -0.31 and -0.18431 values between -0.18 and -0.061620 values between -0.06 and 0.06954 values between 0.06 and 0.19460 values between 0.19 and 0.31119 values between 0.31 and 0.4323 values between 0.43 and 0.557 values between 0.55 and 0.671 values between 0.67 and 0.801 values between 0.92 and 1.042 values between 1.16 and 1.293 values between 1.29 and 1.413 values between 1.41 and 1.531 values between 1.65 and 1.781 values between 1.78 and 1.901 values between 1.90 and 2.021 values between 2.51 and 2.63offnegativepositiveoften

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. 化
  2. brace
  3. strar
  4. 化的
  5. protected
  6. pu
  7. grid
  8. Constraint
  9. constraint
  10. gr

Suppresses

  1. 汲
  2. resident
  3. YLE
  4. ALLE
  5. alley
  6. oman
  7. enames
  8. 蔬
  9. ALLED
  10. poop