Neuron Explorer

Qwen3-0.6B

unclear: words near 'common'

A neuron in layer 20 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 is weak and messy. Its strongest moments tend to fall on words that come right after or near the word 'common', such as 'commonly' or 'common symptoms', but many examples do not fit. Turning it down made the model repeat words like '3D' and 'cold'.

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

    (cooling) coils and circulates it throughout the refrigerator and freezer compartments. Some refrigerators have more than one evaporator fan motor. On refrigerators with only one evaporator, the evaporator is located in the freezer compartment. If the

  2. #2 peak

    taken object-oriented editing the furthest, as the program allows you to open up a clip and access a huge number of clip-related parameters. Sonar, on the other hand, was not designed with object-oriented editing in mind, and takes a

  3. #3 peak

    their conclusion.⏎"Our results suggest that losing just 5 percent or more of one's weight could cut by a quarter to a half the risk for the most common, estrogen-sensitive breast cancers," McTiernan said.<|endoftext|>Is Peru a rich

  4. #4 peak

    the most common pediatric fractures encountered in the emergency department. Supracondylar fractures account for more than 60 percent of elbow fractures, followed by lateral (20%) and medial (

  5. #5 peak

    after the last drink, but can occur days later. Symptoms usually peak by 24 to 72 hours, but may go on for weeks.⏎Common symptoms include:⏎- Anxiety or nervousness⏎- Jumpiness or shakiness⏎

  6. #6 peak

    typical milk snakes with their red, black and yellow rings. As they mature, however, their colors darken until the snakes are uniformly black. Some scientists believe the development of black pigments in this subspecies (type) of milk snake is a

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 a 30000000000000000000000000000000000000000000

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

    consumed all over the world. For example, in the United States and North-Central Europe yogurt is made from cow’s milk, while sheep’s and goat’s milk is used in Turkey. In Egypt and India, yogurt is made from milk of

  2. #2 peak

    show bias than that of women. And scientists in countries with very strong incentives to publish, such as the United States, didn’t seem to have more bias than studies from countries where the pressure was less. These results confirm, with much greater accuracy

  3. #3 peak

    the United States is different from the laetrile/amygdalin produced in Mexico. The patented Laetrile

  4. #4 peak

    33 years in average life span over the past century. A baby born today can expect to live an average of 78 years, compared with 45 years back in 1900. But those gains have not been uniform

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. L19:355 suppresses
  2. L17:2307 suppresses
  3. L15:437 excites
  4. L16:371 excites
  5. L18:210 weakly suppresses
  6. L18:1923 weakly suppresses
  7. L13:1891 weakly suppresses
  8. L16:1181 weakly excites
  9. L16:1007 weakly suppresses
  10. L19:2101 weakly suppresses
  11. L17:2092 weakly suppresses
  12. L13:380 weakly suppresses
  13. L14:223 weakly excites
  14. L19:464 weakly suppresses
  15. L16:2490 weakly suppresses
  16. L18:1037 weakly suppresses
  17. L15:696 weakly suppresses
  18. L18:556 weakly suppresses
  19. L11:562 weakly excites
  20. L16:3059 weakly suppresses
  21. L13:2994 weakly suppresses
  22. L6:883 weakly suppresses
  23. L15:678 weakly suppresses
  24. L15:1079 weakly suppresses

Feeds later layers

  1. L24:104 excites
  2. L22:106 excites
  3. L22:119 weakly excites
  4. L21:170 weakly suppresses
  5. L21:323 weakly excites
  6. L21:98 weakly excites
  7. L27:361 weakly suppresses
  8. L21:843 weakly excites
  9. L22:449 weakly suppresses
  10. L21:324 weakly suppresses
  11. L24:85 weakly suppresses
  12. L27:2995 weakly excites
  13. L24:249 weakly suppresses
  14. L22:29 weakly suppresses
  15. L21:1126 weakly excites
  16. L24:761 weakly excites
  17. L27:1884 weakly excites
  18. L23:348 weakly excites
  19. L25:1612 weakly excites
  20. L23:355 weakly excites
  21. L27:3065 weakly excites
  22. L23:1393 weakly suppresses
  23. L26:486 weakly suppresses
  24. L21:1171 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.

2 values between -7.97 and -7.612 values between -7.24 and -6.882 values between -6.88 and -6.512 values between -6.51 and -6.153 values between -6.15 and -5.784 values between -5.78 and -5.423 values between -5.42 and -5.0518 values between -5.05 and -4.6917 values between -4.69 and -4.3217 values between -4.32 and -3.9616 values between -3.96 and -3.5935 values between -3.59 and -3.2349 values between -3.23 and -2.8688 values between -2.86 and -2.49105 values between -2.49 and -2.13127 values between -2.13 and -1.76210 values between -1.76 and -1.40265 values between -1.40 and -1.03431 values between -1.03 and -0.67571 values between -0.67 and -0.30894 values between -0.30 and 0.06478 values between 0.06 and 0.43206 values between 0.43 and 0.79155 values between 0.79 and 1.1682 values between 1.16 and 1.5251 values between 1.52 and 1.8931 values between 1.89 and 2.2516 values between 2.25 and 2.6215 values between 2.62 and 2.984 values between 2.98 and 3.351 values between 3.35 and 3.722 values between 3.72 and 4.083 values between 4.08 and 4.451 values between 4.45 and 4.811 values between 4.81 and 5.181 values between 6.27 and 6.64offnegativepositiveoften

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. ông
  2. iological
  3. рак
  4. erokee
  5. adget
  6. �建
  7. alph
  8. logistical
  9. _dom
  10. .INTERNAL

Suppresses

  1. | ⏎
  2. ")] ⏎
  3. # ⏎
  4. $ ⏎
  5. ` ⏎
  6. | ⏎
  7. ⏎→ ⏎
  8. ** ⏎
  9. }) ⏎
  10. 恶心