Neuron Explorer

Qwen3-0.6B

year in a written date

A neuron in layer 12 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 year in a written date, such as the 1928 in 'December 7, 1928'. When it was turned up, the model's text drifted off topic and became repetitive, 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

    |⏎|Born||December 7, 1928⏎|Era||20th / 21

  2. #2 peak

    not aware, the draft is a system of conscription that enlists citizens into the military; the system is run using an online sign up form that assigns you a draft number, and should the number be drawn you will be called on for military

  3. #3 peak

    to solve a 2,000 year-old mystery: Who is the Cartographer?More⏎We've sent an email with your order details. Order ID #:⏎

  4. #4 peak

    full Titus Vespasianus Augustus, original name Titus Flavius Vespasianus, (born Dec. 30, 39 ce—died Sept.

  5. #5 peak

    ⏎robert.kern at gmail.com⏎Fri Sep 10 00:10:22 CEST 2010⏎On 9/9/10 4:39 PM, Baba wrote:⏎

  6. #6 peak

    The American historical painter Washington Allston was born in on the Allston Plantation in the Georgetown district of South Carolina Nov. 5, 1779. His parents sent him to Harvard, where not only did

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: 12, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 19, 18, 20

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

    : ||4.4 MB|⏎|Publisher: ||CRC Press|⏎|Date published: || 2015|⏎|ISBN: ||9781498738552 (DRM-PDF)|⏎

  2. #2 peak

    structure = function.⏎(Image from: https://commons.wikimedia.org/w/index.php?curid=1267900)⏎Want to suggest an edit? Have

  3. #3 peak

    ’s website lists 16474 pharmaceutical reports that can be purchased at http://www.fastmr.com/catalog/browse.aspx?category=16&page=1. The online catalog offers a description for each report, which

  4. #4 peak

    12.0 MB|⏎|Publisher: ||Apple Academic Press|⏎|Date published: || 2016|⏎|ISBN: ||9781771882903 (DRM-PDF)|⏎

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. L10:2425 excites
  2. L9:926 weakly suppresses
  3. L1:1058 weakly suppresses closing quote marks after a quoted term
  4. L11:52 weakly excites
  5. L7:1037 weakly excites
  6. L4:2451 weakly suppresses
  7. L9:2453 weakly suppresses
  8. L10:2847 weakly suppresses
  9. L10:987 weakly suppresses
  10. L0:247 weakly excites
  11. L2:2554 weakly excites
  12. L8:2854 weakly suppresses
  13. L10:2832 weakly suppresses
  14. L8:585 weakly suppresses
  15. L6:2018 weakly suppresses
  16. L11:2054 weakly suppresses
  17. L2:2040 weakly excites
  18. L10:354 weakly excites
  19. L10:2394 weakly excites
  20. L7:1381 weakly excites
  21. L9:1246 weakly excites
  22. L8:1049 weakly excites
  23. L11:703 weakly excites
  24. L10:499 weakly excites

Feeds later layers

  1. L22:1076 suppresses
  2. L19:1905 suppresses
  3. L15:2351 suppresses
  4. L16:1676 suppresses
  5. L18:1712 suppresses
  6. L24:654 suppresses
  7. L19:1378 suppresses
  8. L18:2718 suppresses
  9. L13:2602 suppresses
  10. L16:1720 excites
  11. L17:2672 suppresses
  12. L17:932 weakly suppresses
  13. L20:591 weakly suppresses
  14. L16:2196 weakly excites
  15. L19:1033 weakly suppresses
  16. L18:2978 weakly excites
  17. L24:1467 weakly suppresses
  18. L17:3011 weakly excites
  19. L17:2082 weakly suppresses
  20. L16:901 weakly suppresses
  21. L16:2637 weakly suppresses
  22. L15:211 weakly suppresses
  23. L16:2486 weakly excites
  24. L14:1783 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 -2.20 and -2.061 values between -0.82 and -0.689 values between -0.68 and -0.5465 values between -0.54 and -0.40279 values between -0.40 and -0.26872 values between -0.26 and -0.131499 values between -0.13 and 0.01911 values between 0.01 and 0.15220 values between 0.15 and 0.2935 values between 0.29 and 0.439 values between 0.43 and 0.561 values between 0.56 and 0.702 values between 1.12 and 1.261 values between 1.39 and 1.531 values between 1.67 and 1.811 values between 1.95 and 2.081 values between 3.19 and 3.33offnegativepositiveoften

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. طار
  2. orts
  3. 步行
  4. monds
  5. antes
  6. Ladies
  7. ệu
  8. coholic
  9. +(\
  10. ritten

Suppresses

  1. IDs
  2. ID
  3. 번호
  4. 号码
  5. ids
  6. .getID
  7. 号
  8. ID
  9. achi
  10. 编号