Neuron Explorer

Qwen3-0.6B

'a' or 'most' before a quantity

A neuron in layer 18 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 small words like 'a', 'an' or 'most' that come after a verb or 'such' and lead into a noun phrase, as in 'such a long journey' or 'has a lower rate'. Turned up, the model got stuck repeating words like 'the number' or 'the cold'.

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

    . For years, researchers have sought to understand how such a small creature could make such a long journey, particularly because it entails crossing the Gulf of Mexico without stopping.⏎Stories

  2. #2 peak

    50-degree night will slow the plant's growth for some time afterwards.⏎You can start basil seeds indoors, three to four weeks before planting time. Basil has a lower germination rate than many seeds, so plant four to six seeds per pot

  3. #3 peak

    involved. To learn more about workers' compensation options, visit the asbestos workers' compensation section.⏎Additional Information About High Risk Jobs⏎The following list of trades have a

  4. #4 peak

    ?⏎Using different numbers of sticks, how many different triangles are⏎you able to make? Can you make any rules about the numbers of⏎sticks that make the most triangles?⏎Design an arrangement of display boards in

  5. #5 peak

    . This is the version also known as 0.5.0. If you are using an earlier version, you should certainly upgrade; this version fixes quite a few problems, including handling of certain kinds of global variables.⏎The first step in

  6. #6 peak

    to check it every day. In addition, be sure to water it after natural rainfall or if the weather is dry. Succulents in full sunlight should receive a good soaking once a week or so.⏎To find out how often you should be

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 most likely, most effective, and most useful of the most useful of the most useful of the most useful of the most useful of the most useful of the most useful of the most useful of the most useful of the most useful of the

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

    "comprehensive" reforms. In addition to reviewing the overall achievement effects of comprehensive school reform (CSR), they study the specific effects of 29 of the most widely implemented models. They also assess how various CSR components, contextual factors, and

  2. #2 peak

    contagious virus that spreads through direct contact and through the air. It is one of the biggest killers of children worldwide, but can be prevented with two doses of a widely available and inexpensive vaccine.⏎According to the report, published by UNICEF,

  3. #3 peak

    the rain begins to fall on Tanzania’s Tarangire National Park, thousands of zebra, wildebeest and giraffe will begin one of the world’s greatest migrations. But many of the herds trampling across the grass at the

  4. #4 peak

    Council⏎Publication Type: Care guideline<|endoftext|>The Cooper Union for the Advancement of Science and Art, established in 1859, is among the nation's oldest and most distinguished institutions of higher education. The college, founded by inventor,

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. L14:2458 suppresses
  2. L17:3002 suppresses
  3. L13:1729 excites
  4. L17:2263 excites
  5. L14:2220 excites "this" pointing back to a thing
  6. L15:1985 excites
  7. L17:986 suppresses
  8. L17:1377 suppresses indefinite 'a' after a verb
  9. L14:2581 suppresses
  10. L12:1868 weakly suppresses
  11. L17:2756 weakly suppresses
  12. L17:2069 weakly suppresses
  13. L6:2075 weakly suppresses
  14. L12:1218 weakly excites
  15. L14:1337 weakly excites
  16. L14:89 weakly excites
  17. L17:966 weakly suppresses
  18. L13:645 weakly suppresses
  19. L8:74 weakly excites
  20. L15:2590 weakly excites
  21. L17:195 weakly suppresses
  22. L17:1339 weakly excites
  23. L16:1819 weakly excites
  24. L14:1853 weakly suppresses

Feeds later layers

  1. L19:1535 suppresses
  2. L19:2989 suppresses
  3. L19:575 suppresses
  4. L19:1116 suppresses
  5. L20:1006 suppresses
  6. L20:733 suppresses
  7. L21:699 suppresses
  8. L20:382 weakly suppresses
  9. L20:1429 weakly suppresses
  10. L20:89 weakly suppresses
  11. L23:469 weakly suppresses
  12. L20:98 weakly suppresses
  13. L21:2732 weakly excites
  14. L19:1564 weakly excites
  15. L24:372 weakly suppresses
  16. L19:1373 weakly suppresses
  17. L23:511 weakly suppresses
  18. L20:40 weakly excites
  19. L19:870 weakly suppresses
  20. L19:2307 weakly excites
  21. L19:1041 weakly suppresses
  22. L25:135 weakly excites
  23. L19:2106 weakly excites
  24. L20:899 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 -10.37 and -9.921 values between -8.57 and -8.121 values between -7.22 and -6.772 values between -5.87 and -5.421 values between -5.42 and -4.973 values between -4.97 and -4.523 values between -4.52 and -4.072 values between -4.07 and -3.625 values between -3.62 and -3.1711 values between -3.17 and -2.7210 values between -2.72 and -2.276 values between -2.27 and -1.8215 values between -1.82 and -1.3745 values between -1.37 and -0.92243 values between -0.92 and -0.471743 values between -0.47 and -0.021436 values between -0.02 and 0.44223 values between 0.44 and 0.8976 values between 0.89 and 1.3437 values between 1.34 and 1.7913 values between 1.79 and 2.2411 values between 2.24 and 2.6910 values between 2.69 and 3.143 values between 3.14 and 3.593 values between 3.59 and 4.042 values between 4.04 and 4.491 values between 4.49 and 4.941 values between 7.19 and 7.64offnegativepositiveoften

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. ymes
  2. yz
  3. モデ
  4. ym
  5. 当作
  6. (models
  7. ys
  8. _domains
  9. models
  10. }$/

Suppresses

  1. ูล
  2. 能力强
  3. urable
  4. ERN
  5. aturity
  6. lement
  7. ILED
  8. eka
  9. ktion
  10. .HasKey