Neuron Explorer

Qwen3-0.6B

'well' after 'most' or 'and'

A neuron in layer 5 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 weakly on the word 'well' in phrases like 'most well-known' or 'a large and well-defined'. Its activations are small compared with the strongest neurons here. Turning it up did not bring out these phrases in the output.

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

    the offspring of the original polyp remain connected to each other, by a stolonal network or coenosarc. Some species are solitary.⏎While the most well-known zo

  2. #2 peak

    dairy are often considered one of the most important components, and for good reason considering all the essential vitamins and nutrients found in milk. Calcium is one of the most well known ones, but according to Dr. Noreen Iftikhar from

  3. #3 peak

    ." Japanese-speaking listeners, albeit with more variability, were more<|endoftext|>Growth Trends for Related Jobs⏎Physicians and surgeons are some of the highest-paid and most well-educated individuals in the country. According to the Bureau of Labor Statistics, there were

  4. #4 peak

    ELL.)⏎To turn cat-in-pan. To turn traitor, to be a turncoat. The phrase seems to be the French tourner côte en peine (to turn sides in trouble). I do not think it refers to

  5. #5 peak

    on this and that the celery is safe for cats to eat.⏎Cats can definitely eat celery and benefit from it. Cats can eat some veggies in moderate, well, actually, small amounts. Cats can eat celery

  6. #6 peak

    ropical cyclone status, with winds at 40 mph (65 km/h). According to the NHC, that morning the storm maintained a large and well-defined circulation, but had been without organized deep convection for at least 2

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: Let me up there first place: Let me try to get a better look at? Well, encrypted? Maybe it is okay to be able to do with out of the daylight box: Let me be able to do it without the day

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

    on HIV/AIDS, held two high-level briefings as part of a three-part series to expand awareness of the global AIDS agenda outside of the health community, particularly within foreign policy and development circles.⏎World

  2. #2 peak

    proud to share that many of my students will switch to a 100 MIX of addition and subtraction today

  3. #3 peak

    system and how HIV functions. The questions that basic or bench science seek to determine are how a drug might be effective in preventing infection, treating HIV, or curing AIDS. These questions provide researchers with a concept around which they can start to develop a

  4. #4 peak

    of the puzzle in pinpointing the earliest signs of autism,” said Dr. Schultz. “Understanding how autism begins and unfolds in the first years of life will pave the way for more effective interventions and better long-term outcomes for individuals with autism and their

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. L4:949 weakly excites
  2. L4:557 weakly suppresses
  3. L3:1377 weakly suppresses
  4. L3:2437 weakly suppresses
  5. L3:2463 weakly excites
  6. L3:441 weakly excites
  7. L4:1891 weakly suppresses
  8. L2:194 weakly excites
  9. L3:822 weakly excites
  10. L3:1661 weakly suppresses
  11. L4:1090 weakly suppresses
  12. L4:251 weakly suppresses
  13. L2:59 weakly excites
  14. L0:891 weakly excites
  15. L0:1752 weakly excites
  16. L4:2872 weakly suppresses
  17. L2:623 weakly excites
  18. L3:1000 weakly excites
  19. L0:1783 weakly suppresses
  20. L4:838 weakly excites
  21. L0:1820 weakly suppresses
  22. L3:1244 weakly excites
  23. L4:117 weakly suppresses
  24. L4:922 weakly excites

Feeds later layers

  1. L16:592 suppresses
  2. L6:809 suppresses
  3. L17:259 suppresses
  4. L6:2164 suppresses second half of multi-word proper names
  5. L9:929 suppresses 'each' in 'each other'
  6. L6:1198 suppresses
  7. L8:852 suppresses comma inside large numbers
  8. L7:7 suppresses
  9. L9:116 suppresses
  10. L6:2656 weakly suppresses
  11. L6:108 weakly suppresses
  12. L10:1316 weakly suppresses
  13. L6:1100 weakly suppresses
  14. L7:8 weakly suppresses
  15. L6:2069 weakly suppresses
  16. L8:423 weakly suppresses
  17. L20:252 weakly excites
  18. L6:184 weakly suppresses
  19. L6:785 weakly suppresses
  20. L17:1161 weakly suppresses
  21. L7:9 weakly suppresses
  22. L6:35 weakly suppresses
  23. L10:346 weakly suppresses "each" in "each other" phrases
  24. L17:446 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 -1.19 and -1.131 values between -1.13 and -1.071 values between -0.94 and -0.882 values between -0.69 and -0.621 values between -0.62 and -0.564 values between -0.56 and -0.507 values between -0.50 and -0.4310 values between -0.43 and -0.3725 values between -0.37 and -0.3158 values between -0.31 and -0.24156 values between -0.24 and -0.18311 values between -0.18 and -0.12650 values between -0.12 and -0.051307 values between -0.05 and 0.01800 values between 0.01 and 0.07317 values between 0.07 and 0.14129 values between 0.14 and 0.2049 values between 0.20 and 0.2630 values between 0.26 and 0.3317 values between 0.33 and 0.399 values between 0.39 and 0.453 values between 0.45 and 0.527 values between 0.52 and 0.581 values between 0.58 and 0.643 values between 0.64 and 0.713 values between 0.71 and 0.771 values between 0.77 and 0.831 values between 0.83 and 0.901 values between 0.90 and 0.961 values between 0.96 and 1.021 values between 1.02 and 1.091 values between 1.28 and 1.34offnegativepositiveoften

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. rians
  2. .setParameter
  3. Passion
  4. 的意义
  5. .setFont
  6. onyms
  7. 万元以上
  8. yny
  9. HorizontalAlignment
  10. maç

Suppresses

  1. وث
  2. drops
  3. hole
  4. edly
  5. vieille
  6. itta
  7. -window
  8. 隔热
  9. illard
  10. razor