Neuron Explorer

Qwen3-0.6B

the "1" in years 191x

A neuron in layer 20 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 on the digit "1" in years from the 1910s, such as 1917, 1914 or 1925 written as "19 1". It almost never does anything else. Turning it up made the text wander without any link to dates.

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

    all producers, laboratories, dealers, importers, and doctors to register each year to legally buy and sell marijuana.⏎On October 22, 191

  2. #2 peak

    IT disciplines?⏎A frequently-asked question concerns the distinction between CS and Computer Information Systems (CIS). While the specific details of<|endoftext|>In 1919, NYC had 3+ million telephone calls daily — more than the system

  3. #3 peak

    Spain, Britain, the Netherlands, France and Denmark-Norway one after the other. The Islands were purchased by the United States from Denmark in 1917.⏎Saint Thomas is the busiest and the most cosmopolitan of U

  4. #4 peak

    change the world. “I think one of the things we take for granted when we talk about the availability of technology is<|endoftext|>On this date in 1915, what is believed to be the world’s first forest patrol flight was made

  5. #5 peak

    will help reconnect people with the First World War and keep the history and personal recollections of the war alive."⏎The exhibition can be viewed at www.1914.org/faces⏎- More about:⏎- First World War⏎- Imperial

  6. #6 peak

    From Tree Roots⏎- Locate sewer lines – You can call your local public works office or the national 811 “Call Before You Dig” number to locate underground utilities on your property.

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 most important thing that I have to be able to do is to make sure that the same as the same as the world" (111,101). 112,102. 11

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

    Non-alcoholic fatty liver disease typically does not present notable symptoms. To be clear, alcoholic fatty liver disease has few early warning signs, too. Symptoms typically occur once the condition has progressed. Then, in either case, symptoms may be as subtle

  2. #2 peak

    landscape may seem daunting, but by breaking the process down into smaller steps, organizations can leverage their security program to achieve business objectives. One critical component of a solid cybersecurity program is measuring risk.⏎We sat down with senior

  3. #3 peak

    a measles outbreak after it's emerged that one in five children in the county are not fully immunised.⏎It follows a recent

  4. #4 peak

    first year of the survey, 1995, and the second year, 1<|endoftext|>Over the years, the right to recall has come out as one of the most important laws in a democratic country. Although few nations have adopted it

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. L18:499 strongly excites
  2. L17:512 strongly suppresses
  3. L0:871 excites
  4. L4:247 suppresses
  5. L17:360 suppresses
  6. L3:244 suppresses
  7. L19:1615 suppresses
  8. L16:113 suppresses
  9. L10:848 suppresses
  10. L10:1349 excites
  11. L18:420 excites
  12. L18:570 suppresses
  13. L18:474 excites
  14. L14:643 excites
  15. L6:944 suppresses
  16. L15:129 suppresses
  17. L19:2518 excites
  18. L2:2029 suppresses
  19. L11:981 weakly excites
  20. L19:2200 weakly suppresses
  21. L7:674 weakly suppresses
  22. L2:300 weakly suppresses
  23. L0:2471 weakly suppresses
  24. L15:115 weakly excites

Feeds later layers

  1. L25:1965 strongly excites the digit 1 in age ranges
  2. L25:1360 excites
  3. L21:2561 excites
  4. L23:2466 excites
  5. L21:518 excites
  6. L22:2210 excites
  7. L26:728 excites
  8. L27:792 excites
  9. L23:1792 excites
  10. L25:2442 excites
  11. L21:2731 suppresses
  12. L21:2425 weakly suppresses
  13. L25:3022 weakly suppresses
  14. L22:473 weakly suppresses
  15. L27:2252 weakly excites
  16. L27:2850 weakly excites
  17. L23:1228 weakly excites
  18. L25:492 weakly suppresses
  19. L26:1658 weakly excites
  20. L23:2992 weakly suppresses unclear: first pieces of split words
  21. L22:1383 weakly excites
  22. L24:2510 weakly excites
  23. L21:556 weakly suppresses
  24. L26:1638 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.

2 values between -7.71 and -6.132 values between -4.54 and -2.9682 values between -2.96 and -1.373713 values between -1.37 and 0.2147 values between 0.21 and 1.8015 values between 1.80 and 3.388 values between 3.38 and 4.972 values between 4.97 and 6.553 values between 6.55 and 8.144 values between 8.14 and 9.723 values between 11.31 and 12.891 values between 12.89 and 14.481 values between 14.48 and 16.062 values between 16.06 and 17.653 values between 17.65 and 19.233 values between 19.23 and 20.822 values between 20.82 and 22.402 values between 23.99 and 25.571 values between 27.16 and 28.742 values between 28.74 and 30.332 values between 33.50 and 35.081 values between 35.08 and 36.672 values between 36.67 and 38.251 values between 38.25 and 39.841 values between 41.42 and 43.011 values between 46.18 and 47.761 values between 47.76 and 49.351 values between 54.10 and 55.69offnegativepositiveoften

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. st
  2. sti
  3. sts
  4. sto
  5. stro
  6. ˢ
  7. ấy
  8. st
  9. stim
  10. sty

Suppresses

  1. much
  2. th
  3. alach
  4. 不住
  5. odus
  6. _bindings
  7. rows
  8. ical
  9. ounced
  10. uchi