Neuron Explorer

Qwen3-0.6B

blank space before a number

A neuron in layer 24 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 empty space in a phrase like 'a [space] 64-bit' or 'a [space] 35-millimeter', where a number is about to follow. Turning it up made the model write long lists of digits, which fits the same idea.

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

    the garden bed by using a garden fork or tiller to loosen the soil to a depth of 12 to 15 inches, then mix in a 2- to 4-inch layer of compost and/or aged manure.⏎Plant

  2. #2 peak

    the Gemini IV mission during orbit no. 19, on June 4, 1965, with a Hasselblad camera and a 70mm lens. The Gemini IV crew — astronauts Jim McDivitt and

  3. #3 peak

    kinds of combinational circuits are simulated and compared with RALA's. RALA shows the best performance in delay time, area, and power product results. A 64-bit carry-look-ahead adder with RALA is fabricated by

  4. #4 peak

    conducted a study on the movement patterns of round gobies in Muskegon Lake. Using PIT tagged fish we followed 48 round gobies enclosed in a 20x20

  5. #5 peak

    -2 rocket lifted the first ever space-bound camera above New Mexico on a short sub-orbital flight. From an altitude of 65 miles, a 35-millimeter motion picture camera captured grainy monochromatic views of

  6. #6 peak

    depending on where you are. I set up my camera set my lens to its lowest aperture (for my 18mm kit lens it is f/3.5) and then focused

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 early early ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...

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

    along a 100-kilometer long north-south linear vent system over the past 5 million years;

  2. #2 peak

    on the environment as 710,814 flights from London to New York, or eating 16,375,055 kilos of beef.⏎Research shows “a sustained loss of 3 ± 0

  3. #3 peak

    In other words, this is the income that one person needs to meet the cost of basic needs which include the cost of food to provide 2030 kilo calories per person per month, clothing, housing, education, transport, and

  4. #4 peak

    of 15,795 razorback sucker and 11,696 bonytail were released with 134.2 kilohertz (kHz) PIT tags into the Colorado River downstream of Palo Verde Div

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. L23:2142 suppresses
  2. L22:2490 excites
  3. L22:2871 suppresses
  4. L22:2645 suppresses
  5. L18:291 suppresses
  6. L5:222 suppresses
  7. L18:797 excites
  8. L23:2601 suppresses unclear: numbers and distance words
  9. L21:326 suppresses
  10. L17:2446 weakly suppresses
  11. L21:1203 weakly suppresses
  12. L20:2436 weakly suppresses
  13. L22:1574 weakly suppresses
  14. L17:2827 weakly excites
  15. L13:1736 weakly suppresses
  16. L16:1919 weakly suppresses
  17. L22:481 weakly excites
  18. L21:644 weakly suppresses
  19. L21:2146 weakly excites blank before a size number
  20. L14:102 weakly suppresses
  21. L22:2775 weakly excites
  22. L6:1626 weakly suppresses
  23. L23:2791 weakly suppresses
  24. L4:1216 weakly excites

Feeds later layers

  1. L25:253 strongly excites
  2. L26:667 strongly excites
  3. L27:287 strongly excites
  4. L25:46 strongly suppresses
  5. L27:113 strongly suppresses
  6. L26:476 strongly excites
  7. L27:2172 strongly suppresses
  8. L26:2953 strongly suppresses
  9. L27:978 excites
  10. L25:2902 excites
  11. L27:163 suppresses
  12. L26:172 excites
  13. L26:423 suppresses
  14. L26:329 suppresses
  15. L26:611 excites
  16. L25:854 excites
  17. L27:1456 excites
  18. L27:428 excites
  19. L26:991 excites
  20. L27:962 excites
  21. L27:448 suppresses
  22. L27:2089 excites
  23. L26:1127 suppresses
  24. L27:1132 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 -7.04 and -6.452 values between -5.29 and -4.711 values between -4.71 and -4.135 values between -4.13 and -3.556 values between -3.55 and -2.976 values between -2.97 and -2.3811 values between -2.38 and -1.8028 values between -1.80 and -1.22157 values between -1.22 and -0.641318 values between -0.64 and -0.062009 values between -0.06 and 0.52240 values between 0.52 and 1.1053 values between 1.10 and 1.6921 values between 1.69 and 2.2713 values between 2.27 and 2.859 values between 2.85 and 3.437 values between 3.43 and 4.015 values between 4.01 and 4.593 values between 4.59 and 5.174 values between 5.17 and 5.751 values between 6.92 and 7.502 values between 8.08 and 8.662 values between 9.24 and 9.821 values between 9.82 and 10.411 values between 10.41 and 10.991 values between 11.57 and 12.151 values between 15.64 and 16.22offnegativepositiveoften

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. 1
  2. thirty
  3. fifty
  4. 1
  5. 3
  6. 3
  7. 5
  8. seventy
  9. twelve

Suppresses

  1. ipples
  2. cedes
  3. //**⏎
  4. nicas
  5. ANDOM
  6. IGNAL
  7. nds
  8. adow
  9. .uf
  10. pired