Neuron Explorer

Qwen3-0.6B

spot where a number comes next

A neuron in layer 15 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 at the point just before a number is expected, such as the gap in a zip code, a confidence interval, or a number sequence. It is strongly negative on the middle digits of years like 1856 or 1916. When turned up, the model's text did not show a clear theme.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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

    —and it can pass in a flash if you don't have a method for attacking it.<|endoftext|>304 North Cardinal St.⏎Dorchester Center, MA 02124⏎When you’re manually operating a boat, the act

  2. #2 peak

    015 The University of North Carolina Press⏎116 South Boundary Street, Chapel Hill, NC 27514-3808⏎How to Order | Make

  3. #3 peak

    1, 1, 2, 3, 5, 8, 13, 21, 34, 55… each number is the sum ( + ) of its two previous numbers

  4. #4 peak

    996) and abdominal pain (p = 0.018, OR = 0.815, 95% CI = 0.68

  5. #5 peak

    ⏎For more info about recovery and treatment strategy and options, click here⏎626 S. Ferguson Ave. Suite 5⏎Bozeman, MT 59

  6. #6 peak

    tau levels at 24-48 hours post injury (mean difference –0.51 pg/mL, 95% confidence interval, –0.88 to –0.14; P = .008)

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 have a function that is a different version of the same function in the same family of functions in a different version of the solution, I have a function in the same family of a different version of the same function of the same family in

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

    , of the University of Birmingham, and colleagues concluded.⏎Disclosure: The researchers report no relevant financial disclosures.<|endoftext|>You are here⏎Edward Elgar (1857-1934) was 50 years old before he began

  2. #2 peak

    process that people go through when formulating solutions to problems.⏎Herbert Simon was born in Milwaukee, Wisconsin, on June 15th, 1916. He went to the Milwaukee public school system,

  3. #3 peak

    Curtis, George Ticknor, American writer on legal topics, was born at Watertown, Mass., Nov. 28, 1812; and died at New York,

  4. #4 peak

    . A statue of him was erected to commemorate him in London's Parliament Square.⏎Early life and education⏎He was born on 24 May 1870, at the family farm, Bovenplaats, near Malmes

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. L14:542 excites
  2. L11:2554 weakly excites
  3. L14:2846 weakly excites
  4. L4:2337 weakly suppresses
  5. L14:2677 weakly excites
  6. L1:436 weakly suppresses
  7. L7:1333 weakly excites
  8. L6:2866 weakly excites
  9. L2:134 weakly excites
  10. L8:421 weakly excites
  11. L9:1881 weakly excites
  12. L14:1627 weakly excites
  13. L8:3020 weakly excites
  14. L5:2943 weakly excites
  15. L12:107 weakly excites
  16. L14:1700 weakly excites
  17. L6:2093 weakly excites
  18. L14:1829 weakly suppresses
  19. L13:2285 weakly suppresses
  20. L10:288 weakly excites
  21. L7:1609 weakly excites
  22. L6:1320 weakly excites
  23. L2:2639 weakly suppresses
  24. L12:697 weakly suppresses

Feeds later layers

  1. L26:316 weakly excites
  2. L20:1924 weakly suppresses
  3. L17:351 weakly excites
  4. L18:544 weakly excites
  5. L17:2980 weakly excites
  6. L19:672 weakly excites
  7. L16:17 weakly excites
  8. L23:1829 weakly suppresses
  9. L21:349 weakly excites
  10. L17:2496 weakly suppresses
  11. L19:1523 weakly suppresses
  12. L17:119 weakly excites
  13. L19:2376 weakly suppresses
  14. L16:135 weakly excites
  15. L17:328 weakly excites
  16. L20:929 weakly excites
  17. L23:3055 weakly suppresses
  18. L19:673 weakly excites
  19. L21:545 weakly excites
  20. L24:8 weakly suppresses
  21. L23:194 weakly excites
  22. L16:992 weakly excites
  23. L17:784 weakly excites
  24. L22:856 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 -12.28 and -11.911 values between -11.15 and -10.781 values between -10.40 and -10.022 values between -9.65 and -9.271 values between -9.27 and -8.892 values between -8.89 and -8.521 values between -8.52 and -8.144 values between -8.14 and -7.761 values between -7.76 and -7.391 values between -7.01 and -6.631 values between -5.88 and -5.503 values between -5.50 and -5.131 values between -4.75 and -4.371 values between -4.37 and -4.002 values between -4.00 and -3.621 values between -3.62 and -3.252 values between -2.49 and -2.122 values between -2.12 and -1.741 values between -1.74 and -1.366 values between -1.36 and -0.9911 values between -0.99 and -0.61243 values between -0.61 and -0.232793 values between -0.23 and 0.14799 values between 0.14 and 0.5224 values between 0.52 and 0.902 values between 0.90 and 1.271 values between 2.40 and 2.78offnegativepositiveoften

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. 邮箱
  2. 峁
  3. emperature
  4. 둘
  5. çeş
  6. turned
  7. าง
  8. 对方
  9. paged
  10. 율

Suppresses

  1. عني
  2. cı
  3. Corps
  4. uras
  5. 蒟
  6. hypoc
  7. mek
  8. UIS
  9. metaph
  10. uição