Neuron Explorer

Qwen3-0.6B

period after a middle initial

A neuron in layer 6 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 the full stop that follows a single capital letter in a person's name, such as the dot in 'J. Michael Gaziano'. It is negative on the word 'competitive' and related words such as 'advantage'. Turning it down made the model drift toward talk of competitiveness and business strength, though the link is loose.

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

    ammalia⏎About the Author⏎Michael J. Harvey is professor emeritus and an adjunct professor in the Department of Biology at Tennessee Technological University. J. Scott Altenbach is professor emeritus in the Department of

  2. #2 peak

    percent of those people surveyed - including physical therapists, speech and hearing therapists and social workers - were confident in their training in these issues. According to Dr. E. Virginia Laph

  3. #3 peak

    “In the past several years, there’s been an expansion of corn, which has the highest fertilizer per acre … and that’s for biofuels,” said R. Eugene Turner, a Louisiana State University professor who directed the study into the gulf

  4. #4 peak

    the risk of cancer from a typical multivitamin⏎The reduction in total cancers was small but statistically significant, said the study’s lead author, Dr. J. Michael Gaziano, a cardiologist at Brigham and Women’s Hospital and the VA

  5. #5 peak

    taking their pills as directed, the study found.⏎“It does appear that there is a modest reduction in cancer among middle-aged and older men,” said Dr. J. Michael Gaziano, chief of the division of aging at Brigham and Women’s Hospital

  6. #6 peak

    so the pain becomes more manageable.⏎Mindfulness-Based Cognitive Therapy grew largely from Jon Kabat-Zinn's work. Zindel V. Segal, J.

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 potential of a leader in a global economy. The key to success in an increasingly competitive global economy, the ability to compete globally, and the potential of a leader in a competitive, global, and prosperous business world. The ability 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

    a position of strength thanks to technological turning points taken early enough and, above all, thought out so as to constitute a competitive advantage. One can take the example of choosing nuclear power to ensure a massive and relatively

  2. #2 peak

    United States to remain globally competitive we need 60-percent of our population to obtain a post-secondary degree by

  3. #3 peak

    measure the strength of a chemical bond between an antibody and a virus particle.⏎These and other extreme capabilities are key to the nation’s high-technology future, the competitiveness of its industries, and the health and well-being of its citizens. They are

  4. #4 peak

    loans and connections he built to buy out his competitors. This enabled him to build his empire.⏎3. Rockefeller made deals with the railroads to give him an edge over his competitors and allow him to price his oil cheaper then theirs. Thinking outside

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. L3:2931 weakly suppresses
  2. L2:2057 weakly suppresses
  3. L5:215 weakly excites
  4. L5:1994 weakly suppresses
  5. L2:2692 weakly excites
  6. L4:2947 weakly excites
  7. L1:534 weakly excites
  8. L3:115 weakly suppresses
  9. L1:2581 weakly excites
  10. L2:547 weakly suppresses
  11. L2:2976 weakly excites
  12. L5:2603 weakly excites
  13. L3:1901 weakly suppresses
  14. L2:1959 weakly suppresses
  15. L5:193 weakly excites
  16. L5:2143 weakly excites
  17. L4:1427 weakly suppresses
  18. L5:158 weakly suppresses
  19. L5:60 weakly excites
  20. L1:2890 weakly excites
  21. L5:2116 weakly suppresses
  22. L4:1405 weakly suppresses
  23. L1:2305 weakly excites
  24. L4:1495 weakly suppresses

Feeds later layers

  1. L22:2738 suppresses
  2. L11:1284 suppresses
  3. L17:3015 suppresses
  4. L15:2566 suppresses
  5. L10:2729 suppresses
  6. L7:1462 suppresses
  7. L11:1108 weakly suppresses
  8. L9:1937 weakly suppresses
  9. L11:1986 weakly suppresses
  10. L11:2213 weakly suppresses
  11. L17:1354 weakly suppresses
  12. L9:2002 weakly suppresses
  13. L12:3049 weakly excites
  14. L21:1065 weakly suppresses
  15. L7:2662 weakly suppresses
  16. L20:2748 weakly suppresses
  17. L11:108 weakly suppresses
  18. L7:2451 weakly suppresses
  19. L9:150 weakly suppresses
  20. L9:3047 weakly suppresses
  21. L7:2436 weakly suppresses
  22. L11:2003 weakly excites
  23. L14:2680 weakly excites
  24. L8:2364 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 -2.14 and -2.051 values between -1.67 and -1.581 values between -1.58 and -1.481 values between -1.10 and -1.011 values between -1.01 and -0.921 values between -0.92 and -0.823 values between -0.82 and -0.735 values between -0.73 and -0.634 values between -0.63 and -0.549 values between -0.54 and -0.4434 values between -0.44 and -0.35114 values between -0.35 and -0.26293 values between -0.26 and -0.16711 values between -0.16 and -0.071277 values between -0.07 and 0.03834 values between 0.03 and 0.12396 values between 0.12 and 0.22147 values between 0.22 and 0.3146 values between 0.31 and 0.4112 values between 0.41 and 0.504 values between 0.50 and 0.602 values between 0.60 and 0.694 values between 0.69 and 0.781 values between 0.78 and 0.883 values between 0.88 and 0.971 values between 1.16 and 1.261 values between 1.26 and 1.351 values between 1.54 and 1.63offnegativepositiveoften

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. .rawValue
  2. iram
  3. imuth
  4. .hist
  5. indeed
  6. .setOnAction
  7. numberOfRows
  8. OperationException
  9. ansch
  10. odor

Suppresses

  1. leans
  2. 优势
  3. naissance
  4. advantage
  5. золот
  6. 有利
  7. 竞争优势
  8. ogne
  9. 鹬
  10. playing