Neuron Explorer

Qwen3-0.6B

quantities right after 'a' in numbers

A neuron in layer 4 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 word just before a number or amount, especially 'a' or 'a 20%' style phrases, such as 'a 100-gram serving' or 'a 5 minute shower'. When turned up, the text drifted into rambling about numbers and calculations, with no clear single theme.

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

    , a nutrient your body needs to repair cells and build new ones.⏎Source of Minerals⏎Tamales contain good amounts of iron and phosphorus. Eating a 100-gram serving of tamales will provide you with 1

  2. #2 peak

    the daily value for saturated fat. Eating too much total fat and saturated fat elevates your blood cholesterol levels and increases your risk for heart disease. Also, a 100-gram serving of tamales delivers approximately 3.1

  3. #3 peak

    2.7 m3 per MWh, but wind power uses only a fraction of these amounts.⏎In the US, the Department of Energy estimates that with a 20% share of wind power in

  4. #4 peak

    low carbohydrate diet were at higher risk of premature death due to stroke, coronary heart disease and cancer. It was found that people with lowest intake of carb had a 32 percent higher risk of all-cause death. Besides, death risk from

  5. #5 peak

    using less than 10 percent of its surface area and supply an additional six to eight hours of power per day. Modeling by Sunengy shows that a 240 MW LSA system could increase annual energy generation at the Portuguese hydro

  6. #6 peak

    a shower instead of a bath – a bath can use 15 – 25 gallons of hot water compared to the 10 gallons used for a 5 minute shower.⏎- Lower the thermostat setting on your water heater.⏎- Only

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: I have been thinking about lately You have been thinking about lately So you can use to You can use to So you can use to You can use to You can use to You can use to You can

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

    brushed and brushless DC motors. Typical loaded maximum operating speeds for stepper motors are around 1000 RPM.⏎- Complex electronic controls are required.⏎Figure 1: Structure of motors.⏎Types of stepping motors⏎There are several basic types

  2. #2 peak

    machine with network file sharing enabled.⏎- Visualize links, by clicking on the visualize link the user interface creates a hierarchical animated tree visualization of trace links. Figure 1 below shows the user interface.⏎This capstone project has been a tremendous learning

  3. #3 peak

    :⏎- Interpersonal, project management, risk management and time management skills.⏎- Software development in Java using various web-development technologies.⏎Figure 1: Traceability User Interface⏎Back to top<|endoftext|>William Gilly has seen

  4. #4 peak

    .⏎a. Self-equalizing Anchor. A self-equalizing anchor will maintain an equal load on each individual point as the direction of pull changes (Figure 5-18). This is sometimes used in rappelling when the route must

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. L3:115 suppresses
  2. L3:464 weakly excites
  3. L3:3 weakly suppresses
  4. L3:456 weakly suppresses
  5. L3:421 weakly suppresses
  6. L3:286 weakly excites
  7. L3:2825 weakly suppresses
  8. L3:187 weakly excites
  9. L2:2978 weakly suppresses
  10. L2:1998 weakly excites
  11. L0:2371 weakly excites
  12. L2:493 weakly suppresses
  13. L3:59 weakly excites
  14. L2:111 weakly suppresses
  15. L3:1243 weakly excites
  16. L0:706 weakly suppresses
  17. L2:1350 weakly excites
  18. L1:1530 weakly suppresses
  19. L3:591 weakly suppresses
  20. L3:2517 weakly excites
  21. L2:186 weakly excites
  22. L1:108 weakly suppresses
  23. L0:17 weakly excites
  24. L3:2603 weakly excites

Feeds later layers

  1. L6:908 suppresses
  2. L10:1840 excites
  3. L16:3022 excites
  4. L11:938 weakly suppresses
  5. L13:2524 weakly suppresses
  6. L14:549 weakly excites blank spot right before a number
  7. L23:297 weakly excites
  8. L11:2846 weakly suppresses
  9. L21:175 weakly excites
  10. L11:836 weakly excites
  11. L10:435 weakly excites
  12. L19:2877 weakly suppresses
  13. L27:1078 weakly excites
  14. L12:2515 weakly suppresses
  15. L19:2870 weakly suppresses
  16. L17:553 weakly excites
  17. L5:533 weakly excites
  18. L24:2854 weakly suppresses
  19. L12:696 weakly excites
  20. L24:1204 weakly suppresses
  21. L6:442 weakly suppresses
  22. L8:1707 weakly excites
  23. L19:1614 weakly excites
  24. L16:2866 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 -1.57 and -1.451 values between -1.33 and -1.221 values between -0.99 and -0.872 values between -0.87 and -0.761 values between -0.76 and -0.642 values between -0.64 and -0.532 values between -0.53 and -0.4140 values between -0.41 and -0.29207 values between -0.29 and -0.18684 values between -0.18 and -0.061676 values between -0.06 and 0.05912 values between 0.05 and 0.17248 values between 0.17 and 0.2853 values between 0.28 and 0.4011 values between 0.40 and 0.525 values between 0.52 and 0.635 values between 0.63 and 0.753 values between 0.75 and 0.866 values between 0.86 and 0.983 values between 0.98 and 1.093 values between 1.09 and 1.212 values between 1.21 and 1.327 values between 1.32 and 1.441 values between 1.44 and 1.558 values between 1.55 and 1.674 values between 1.67 and 1.795 values between 1.79 and 1.902 values between 1.90 and 2.021 values between 2.02 and 2.131 values between 2.13 and 2.251 values between 2.25 and 2.362 values between 2.36 and 2.483 values between 2.48 and 2.594 values between 2.71 and 2.831 values between 2.94 and 3.06offnegativepositiveoften

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. CLUSIVE
  2. baseline
  3. onymous
  4. thr
  5. 款车型
  6. 高的
  7. baseline
  8. AccessException
  9. HostException
  10. .hxx

Suppresses

  1. úa
  2. "`
  3. Main
  4. Teach
  5. Instead
  6. -mail
  7. Keep
  8. sb
  9. {:
  10. ssa