Neuron Explorer

Qwen3-0.6B

the word 'more' after 'a'

A neuron in layer 24 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 reacts to the word "more" when it comes after "a" or "a form of" in phrases like "a more serious problem" or "a more perfect union". When it was turned up, the model kept repeating words like "perfect" and "numerous", and when turned down it repeated "more" and "than".

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

    in Europe. The Commission notes that the main environmental threat from bio-waste is the production of methane, a greenhouse gas (GHG) 25 times more potent than carbon dioxide.⏎In the Communication, the Commission recommends realizing bio-waste

  2. #2 peak

    nose are fragile, so it is no wonder that they can easily bleed. However, sometimes nose bleeds can become frequent. This may be an indication of a more serious problem.⏎One of the most common causes of frequent nose bleeds is dry

  3. #3 peak

    a crack in the terrace. As long as water flows up this point of weakness in the terrace, the cone continues to grow, but if the water finds a more convenient underground channel or if the orifice of the spring is sealed over by trav

  4. #4 peak

    and white rice, cause a rapid increase in blood sugar levels. On the other side, foods with low GI, such as whole grains and sweet potato, are more slowly digested, causing a lower and gentler fluctuation of sugar levels in

  5. #5 peak

    set forth in the “preamble,” or opening clause, of the Constitution of the United States;⏎1. “To form a more perfect union;”⏎2. “To establish justice;”⏎3. “To

  6. #6 peak

    causes of bedwetting in elderly adults. At its least, adult bedwetting is embarrassing and inconvenient; at its worst, it can be indicative of a more serious urinary tract problem. To help your health care provider properly determine a reason and

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: more than than than than than than than than than than than than than than than than than than than than than than than than than than than than than than than than than than than than than than than than than than than than than than than

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

    that you select: .⏎You can also select a web site from the following list:⏎Select the China site (in Chinese or English) for best site performance. Other MathWorks country sites are not optimized for visits from your location.⏎Contact your local

  2. #2 peak

    , and was incorporated into the Kingdom of Siam in 1786. Patani was divided into the three provinces Patani, Yala, and Narathiwat in 1808. Thailand's influence over the region

  3. #3 peak

    ÷ 4 = 3,125 lbs⏎According to this calculation, the vehicle mentioned above will require each wheel to have a safe working load greater than 3,125 lbs.⏎Other factors that can affect wheel performance

  4. #4 peak

    coastal cliffs and villages and towns.<|endoftext|>Have you taken nonsteroidal anti-inflammatory drugs (NSAIDs) like diclofenac, ibuprofen or naproxen? A study of soldiers showed that people taking high doses of NSAIDs

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. L18:3065 excites
  2. L23:2405 excites
  3. L21:2777 excites
  4. L4:344 suppresses
  5. L18:2374 suppresses
  6. L20:212 weakly excites
  7. L17:2447 weakly suppresses
  8. L7:2825 weakly excites
  9. L16:588 weakly suppresses
  10. L22:935 weakly suppresses
  11. L1:2396 weakly excites
  12. L22:2527 weakly excites
  13. L14:1426 weakly suppresses
  14. L15:699 weakly excites
  15. L4:1873 weakly suppresses
  16. L13:1276 weakly suppresses
  17. L19:739 weakly excites
  18. L3:1316 weakly suppresses
  19. L8:1572 weakly suppresses
  20. L22:1711 weakly suppresses
  21. L4:656 weakly excites
  22. L22:1046 weakly excites
  23. L23:3034 weakly excites
  24. L5:1087 weakly suppresses

Feeds later layers

  1. L26:551 suppresses
  2. L27:948 suppresses
  3. L26:1908 suppresses
  4. L27:755 suppresses
  5. L25:1196 weakly suppresses
  6. L27:1896 weakly suppresses split names and comparative stems
  7. L27:1006 weakly suppresses
  8. L27:1734 weakly suppresses
  9. L25:1425 weakly suppresses
  10. L26:1597 weakly suppresses
  11. L25:2908 weakly excites
  12. L26:880 weakly suppresses
  13. L27:1451 weakly excites
  14. L27:545 weakly excites
  15. L25:1046 weakly suppresses
  16. L25:2986 weakly suppresses '1' in 21st or 11th, before 'st'
  17. L25:787 weakly excites
  18. L26:1007 weakly suppresses
  19. L25:1185 weakly suppresses
  20. L25:2371 weakly suppresses
  21. L26:1382 weakly suppresses
  22. L26:396 weakly suppresses
  23. L27:1964 weakly suppresses
  24. L25:3053 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 -14.41 and -11.511 values between -11.51 and -8.613 values between -8.61 and -5.717 values between -5.71 and -2.802585 values between -2.80 and 0.101276 values between 0.10 and 3.0015 values between 3.00 and 5.904 values between 5.90 and 8.813 values between 8.81 and 11.711 values between 11.71 and 14.611 values between 14.61 and 17.512 values between 20.42 and 23.321 values between 32.03 and 34.931 values between 37.83 and 40.732 values between 43.64 and 46.541 values between 46.54 and 49.441 values between 63.95 and 66.861 values between 78.47 and 81.372 values between 98.79 and 101.69offnegativepositiveoften

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. numerous
  2. Every
  3. ",&
  4. Every
  5. ropa
  6. )],
  7. adamente
  8. %=
  9. ้ม
  10. ้อง

Suppresses

  1. 更
  2. more
  3. 更有
  4. 更为
  5. more
  6. hơn
  7. 更加
  8. 更大的
  9. 更大
  10. più