Neuron Explorer

Qwen3-0.6B

measurement words before a value

A neuron in layer 25 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on words that name a size or measurement, such as 'waist' in waist circumference, 'correct' in correct temperature, or 'body' in body length. It seems to expect a number or unit to follow. Turning it up made the replies wander into numbers and sizes, though the text got muddled.

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

    of novelties. Sparklers can burn at more than 1,000 degrees Fahrenheit and should be used with caution.⏎• Cook food to the correct temperature and refrigerate uneaten food promptly.⏎• Wash/rinse fruits and

  2. #2 peak

    measure physical activity and sedentary time. They also used questionnaires to assess the children’s screen and sleep time. The researchers measured weight, height, and waist circumference and used dual-energy x-ray absorptiometry to assess body composition.⏎

  3. #3 peak

    -dwelling marine fish. They live in coastal waters over sandy and muddy substrates at depths ranging from a few metres to 150 metres.⏎Maximum body length 15 cm, usually up to 10 cm. Maximum weight

  4. #4 peak

    The Total Wellbeing Diet put out by the Australian CSIRO, Metabolic Syndrome X should be suspected if you have three of more of the following conditions:⏎..... Waist circumference bigger than 88cm (34 inches)⏎..... HDL cholesterol

  5. #5 peak

    .001. See also millimeters into meters calculator. See also millimeters into meters calculator.⏎Meter (abbreviated m) – is the basic unit of length in systems such as: SI, MKS, or MTS. The

  6. #6 peak

    When it comes to poultry, most of us think of the Thanksgiving turkey that comes with a pop-up thermometer. Any and all poultry you cook should be cooked to an internal temperature of 165 degrees Fahrenheit. Another study related to meat thermometer

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 average of the odds ratio and the odds ratio area. I mean, odds ratio area is the sum of the odds ratio volumes, and the odds ratio is the area of the odds ratio area. So, the odds ratio is the total

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

    pressure, and it becomes harder to draw out the remaining crude. Engineers can try to add pressure underground by putting in seawater or carbon dioxide. While this increases the field’s output, it also sours the product and leads to more pipeline corrosion

  2. #2 peak

    of the clock system connected to properties such as plant size and stress tolerance, with significant potential benefits to agriculture.⏎For more information, please contact:⏎Dr. Norihito Nakamichi⏎Biodynamics Research Team⏎RIKEN Plant Science Center⏎

  3. #3 peak

    ational hygienist Kate Cole specializes in workplace safety and has provided advice on ventilation during the pandemic.⏎âWhen we have people in indoor spaces, it creates a high risk environment because people generate aerosols,â she said.⏎Canberra sisters Krist

  4. #4 peak

    rons to maintain a stable GFR even when arterial blood pressure changes.⏎Which mechanism of renal autoregulation responds to an INCREASE in BP which causes an increased STRETCH in the afferent arteriole wall, causing smooth muscle

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. L19:603 strongly suppresses
  2. L22:557 strongly suppresses
  3. L23:142 strongly suppresses
  4. L22:2648 strongly excites
  5. L19:2761 strongly suppresses
  6. L19:1603 strongly excites
  7. L22:190 strongly excites
  8. L21:599 strongly excites
  9. L23:455 suppresses
  10. L22:420 excites
  11. L20:425 suppresses
  12. L18:1447 excites
  13. L18:2049 excites
  14. L20:462 excites
  15. L15:1423 suppresses
  16. L16:2258 suppresses
  17. L18:1215 excites
  18. L18:2157 excites
  19. L19:1233 excites
  20. L20:2213 suppresses
  21. L18:1555 excites
  22. L23:3022 excites
  23. L20:1284 suppresses
  24. L20:1800 suppresses

Feeds later layers

  1. L27:758 strongly suppresses
  2. L26:181 suppresses
  3. L27:2830 suppresses
  4. L27:2171 suppresses
  5. L27:3014 suppresses
  6. L26:528 suppresses
  7. L26:861 suppresses
  8. L27:974 weakly suppresses
  9. L26:2464 weakly suppresses
  10. L27:1404 weakly suppresses
  11. L26:2305 weakly suppresses
  12. L26:910 weakly suppresses
  13. L26:2180 weakly suppresses
  14. L27:2596 weakly suppresses
  15. L27:1926 weakly suppresses
  16. L26:2136 weakly suppresses
  17. L27:2792 weakly suppresses
  18. L26:1508 weakly suppresses
  19. L26:1663 weakly suppresses
  20. L26:1219 weakly suppresses
  21. L27:1395 weakly suppresses
  22. L27:1304 weakly suppresses
  23. L27:2815 weakly suppresses
  24. L27:1488 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 -16.55 and -15.911 values between -14.00 and -13.362 values between -13.36 and -12.722 values between -12.72 and -12.093 values between -10.18 and -9.541 values between -9.54 and -8.904 values between -8.90 and -8.262 values between -8.26 and -7.633 values between -7.63 and -6.996 values between -6.99 and -6.3517 values between -6.35 and -5.7212 values between -5.72 and -5.0811 values between -5.08 and -4.4419 values between -4.44 and -3.8135 values between -3.81 and -3.1752 values between -3.17 and -2.5372 values between -2.53 and -1.89106 values between -1.89 and -1.26189 values between -1.26 and -0.62536 values between -0.62 and 0.021739 values between 0.02 and 0.661034 values between 0.66 and 1.2944 values between 1.29 and 1.935 values between 2.57 and 3.203 values between 3.20 and 3.841 values between 3.84 and 4.481 values between 4.48 and 5.122 values between 5.12 and 5.752 values between 5.75 and 6.392 values between 7.03 and 7.661 values between 8.30 and 8.94offnegativepositiveoften

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. zig
  2. rending
  3. iert
  4. usercontent
  5. ilians
  6. 缨
  7. felt
  8. vey
  9. azine
  10. akening

Suppresses

  1. 量
  2. rate
  3. volume
  4. distance
  5. 面积
  6. 体积
  7. rates
  8. 比例
  9. 总量
  10. duration