Neuron Explorer

Qwen3-0.6B

'most' or 'real' before the key point

A neuron in layer 13 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 like 'most', 'real', 'more' or 'crucial' that point to the single most important thing in a sentence, as in 'the most important component' or 'the real benefits'. Turning it up made the text keep saying 'the most important' and 'the most exciting'.

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

    scene allows mapping of land use along the coastal plain, including the boundaries of the Shengli Oil Field, China’s second largest oil field. But perhaps the most interesting aspect of this image is the large patch of green

  2. #2 peak

    the Bible servants or advisors are mentioned without being named. Maybe we could find their names elsewhere in Persian history if we took time to look.⏎But maybe the more important question is why did it take 7 men to go and bring one wife to

  3. #3 peak

    First, let’s review why fruits and veggies are so good for us. Sure, we know carrots help our eyesight and bananas give us potassium. But the real benefits of produce are gleaned from studies that have examined populations of people who have

  4. #4 peak

    an entire basic computer fitted on a single chip . Sure, computers purchased today usually come with peripherals like monitors, hard drives, and DVD drives, but the most important component of the system is the microprocessor. You most likely have heard of companies

  5. #5 peak

    with an "Arsenal of Democracy".⏎Famed British author Norman Davies noted that American forces fought with distinction on two fronts, but that it was our home factories that turned the tide. In the late 1930's, American industry

  6. #6 peak

    the CSI protocol, quickly at first (color was a concept they understood) and then more slowly, defining symbol and image and the difference between the two. More important then their actual choice of color, symbol and image, was their explanation for each

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 a student of the house, and it is a house in the circle of the square, and it is a circle, and the square of a circle, for example, in the square of a circle for the square of a

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

    : C F O S U.⏎No direct anagrams for focus found in this word list.⏎List shorter words within focus, sorted by length⏎All words formed from focus by changing one letter⏎Browse words starting with focus by next letter<|endoftext|>I

  2. #2 peak

    Kennedy at Rice University in Houston in September of 1962:⏎“We choose to go to the Moon in this decade and do the other things, not because they are easy, but because they are hard, because

  3. #3 peak

    /gifted student educator with Brevard Public Schools. ...<|endoftext|>When planning your next landscape project, consider using native species. Native plants tend to grow better than introduced species because they have evolved under local growing conditions. Native plants are less

  4. #4 peak

    become dislodged and shift out of place, which causes the sensation of vertigo. The condition occurs most often in people 50 and older and is more common in women than in men.⏎Doctors treat BPPV by

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. L12:1484 excites
  2. L8:1074 excites
  3. L12:1548 excites
  4. L10:1063 weakly suppresses
  5. L4:2207 weakly suppresses
  6. L4:1354 weakly excites
  7. L4:344 weakly suppresses
  8. L9:2477 weakly suppresses
  9. L3:1316 weakly suppresses
  10. L11:490 weakly excites
  11. L9:1046 weakly excites
  12. L11:2964 weakly suppresses
  13. L7:1986 weakly suppresses
  14. L10:1058 weakly suppresses
  15. L6:2604 weakly suppresses
  16. L11:821 weakly excites
  17. L9:1896 weakly suppresses
  18. L9:2743 weakly excites
  19. L11:2101 weakly excites
  20. L8:2380 weakly suppresses
  21. L5:613 weakly suppresses
  22. L5:1440 weakly excites
  23. L8:2298 weakly suppresses
  24. L8:2196 weakly suppresses comparison words like more, same, but

Feeds later layers

  1. L14:1869 excites
  2. L15:2523 excites
  3. L15:2849 excites
  4. L14:2240 excites
  5. L18:2398 excites
  6. L27:1086 excites
  7. L17:2628 excites
  8. L15:2340 excites
  9. L17:2707 weakly excites
  10. L15:1470 weakly excites
  11. L22:969 weakly excites
  12. L18:2761 weakly excites
  13. L16:2647 weakly excites
  14. L23:2491 weakly excites
  15. L14:1795 weakly excites
  16. L20:1076 weakly excites
  17. L16:2833 weakly excites
  18. L19:2597 weakly excites
  19. L18:1553 weakly excites
  20. L26:1525 weakly excites
  21. L15:384 weakly excites
  22. L14:2532 weakly excites
  23. L15:3061 weakly excites
  24. L15:2637 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 -1.42 and -1.273 values between -0.98 and -0.843 values between -0.84 and -0.695 values between -0.69 and -0.5425 values between -0.54 and -0.40125 values between -0.40 and -0.25541 values between -0.25 and -0.111727 values between -0.11 and 0.04999 values between 0.04 and 0.18328 values between 0.18 and 0.3396 values between 0.33 and 0.4821 values between 0.48 and 0.626 values between 0.62 and 0.776 values between 0.77 and 0.913 values between 0.91 and 1.062 values between 1.06 and 1.203 values between 1.20 and 1.351 values between 1.35 and 1.501 values between 1.50 and 1.642 values between 1.64 and 1.792 values between 1.79 and 1.931 values between 1.93 and 2.081 values between 2.08 and 2.221 values between 2.22 and 2.371 values between 2.37 and 2.511 values between 2.51 and 2.661 values between 2.66 and 2.811 values between 3.68 and 3.821 values between 4.26 and 4.41offnegativepositiveoften

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. most
  2. 最有
  3. 最具
  4. 最
  5. 最为
  6. 最受欢迎
  7. onyms
  8. 最关键
  9. 最高
  10. primary

Suppresses

  1. IENTATION
  2. ampp
  3. acie
  4. falls
  5. pector
  6. Vari
  7. ousse
  8. ))/
  9. ))^
  10. uable