Neuron Explorer

Qwen3-0.6B

the zero inside a year

A neuron in layer 22 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 second digit of a year, mostly the "0" in numbers like 2009, 2001 or 1901. When turned down, the model got stuck repeating the word "should", which matches what this neuron tends to suppress.

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

    : “There is no reason for a landscape ash tree to die from emerald ash borer anymore.”⏎The destructive insect, first detected here in 2009, has been found in Hennepin

  2. #2 peak

    nature of what – surface, slush or seas – lay beneath, Titan continues to serve up unending surprises and mysteries. Since the summer of 2009, the tilt of its orbit around Saturn has caused its northern and southern

  3. #3 peak

    The following article refers to President Bush's deliberations about allowing funding for stem cell research. On August 9, 2001, Bush gave his qualified approval for such funding, permitting it for research done

  4. #4 peak

    other regulations, websites or to contact your legislators.⏎ADA Anniversary Tool Kit includes: Overview of the Americans with Disabilities Act and the ADA Amendments Act of 2008; ADA Findings, Purpose, and History; The ADA from a

  5. #5 peak

    1%) as influenza B. Of the sub-typed influenza A viruses reported, 51.8% were influenza A(H1N1)2009 and 48% were influenza A(H3N2).⏎(Flu

  6. #6 peak

    shots from a .32-caliber Iver-Johnson revolver to assassinate William McKinley. He was shot on September 6, 1901 in the late afternoon by Leon Czolgosz during a public hand

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: Should I should I should should should should should should should should should should should should should should should should should should should should should should should should should should should should should should should should should should should should should should should should should should should should

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

    at the bottom left of the screen and click on it. It should open up after clicking.⏎· Type HydroShare.org into the browser and click on it – it should open up.⏎·

  2. #2 peak

    the Outlook window, choose the transferred particulars folder.⏎- Then open the email that you want to recall. You must double-click to open the communication. The Recall option isn’t available for a communication displayed in the Reading Pane.⏎- Now from the

  3. #3 peak

    2: Watch the child try to stand up with help from the furniture from just before the end of their first year, and finally walk with some help. They will bring objects to you, having built their confidence enough to carry things and interact

  4. #4 peak

    0.⏎- Inside the loop, we are printing the element for index i and decrementing the value of i by 1.⏎If you run this program, it will print the below output:⏎5 4 3 2 1⏎

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. L20:2748 strongly suppresses
  2. L19:2624 strongly suppresses
  3. L21:2363 excites
  4. L21:2025 suppresses
  5. L19:2748 suppresses
  6. L18:2120 suppresses
  7. L21:2121 excites
  8. L21:1826 weakly excites
  9. L15:2327 weakly suppresses
  10. L7:1963 weakly excites
  11. L11:1372 weakly excites
  12. L21:1913 weakly suppresses
  13. L19:1678 weakly excites
  14. L18:2959 weakly excites
  15. L2:611 weakly suppresses
  16. L19:1700 weakly suppresses
  17. L16:2197 weakly excites
  18. L18:2036 weakly excites
  19. L5:2380 weakly suppresses
  20. L15:1468 weakly suppresses
  21. L17:1493 weakly suppresses
  22. L15:2076 weakly suppresses
  23. L4:2231 weakly suppresses
  24. L20:3048 weakly excites

Feeds later layers

  1. L27:1236 strongly suppresses
  2. L24:764 strongly suppresses
  3. L23:1876 suppresses
  4. L24:2481 suppresses
  5. L24:1897 suppresses
  6. L27:2484 suppresses
  7. L25:959 weakly suppresses
  8. L25:1181 weakly suppresses
  9. L23:2700 weakly suppresses
  10. L23:1173 weakly suppresses
  11. L25:2905 weakly suppresses
  12. L27:1867 weakly excites
  13. L26:1933 weakly suppresses
  14. L24:1571 weakly excites
  15. L23:1993 weakly suppresses
  16. L26:1276 weakly suppresses
  17. L27:2213 weakly excites
  18. L26:1494 weakly excites
  19. L24:471 weakly excites
  20. L25:2802 weakly excites
  21. L23:2104 weakly excites
  22. L25:695 weakly excites
  23. L25:2145 weakly suppresses
  24. L26:2014 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 -4.86 and -4.591 values between -4.31 and -4.031 values between -4.03 and -3.751 values between -3.47 and -3.191 values between -2.92 and -2.642 values between -2.36 and -2.082 values between -2.08 and -1.804 values between -1.80 and -1.525 values between -1.52 and -1.2511 values between -1.25 and -0.9720 values between -0.97 and -0.6989 values between -0.69 and -0.41322 values between -0.41 and -0.131329 values between -0.13 and 0.151150 values between 0.15 and 0.43597 values between 0.43 and 0.70246 values between 0.70 and 0.9888 values between 0.98 and 1.2620 values between 1.26 and 1.544 values between 1.54 and 1.824 values between 1.82 and 2.102 values between 2.10 and 2.371 values between 2.37 and 2.651 values between 2.65 and 2.931 values between 3.49 and 3.771 values between 4.32 and 4.601 values between 4.60 and 4.881 values between 5.16 and 5.441 values between 5.71 and 5.991 values between 5.99 and 6.27offnegativepositiveoften

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. sätze
  2. meet
  3. apixel
  4. atur
  5. met
  6. uzzi
  7. él
  8. meet
  9. omet
  10. 样子

Suppresses

  1. should
  2. should
  3. 应该
  4. Should
  5. 应
  6. shouldn
  7. 应当
  8. Should
  9. .should
  10. 應該