Neuron Explorer

Qwen3-0.6B

'of' after group or combination nouns

A neuron in layer 1 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 word 'of' when it follows words like 'team', 'combination' or 'process', as in 'a team of scientists'. Turning it up did not produce a clear change in the story the model wrote.

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

    extensive use of electric motors, steam engines or water turbines often supplied the power for factory machinery. One engine or turbine might run a whole factory through a complicated system of shafts, pulleys, and belts. Pulleys for flat<|endoftext|>Today is

  2. #2 peak

    and monkey dialects⏎The same species of monkeys located in separate geographic regions use their alarm calls differently to warn of approaching predators, a linguistic analysis by a team of scientists reveals. The study, which appears in the journal Linguistics and Philosophy,

  3. #3 peak

    offer huge potential to deliver fast, powerful and more efficient calculations than traditional computers, experts say.⏎Recent progress in the design and engineering of quantum computers led a team of scientists from the Universities of Edinburgh and Copenhagen to investigate if they could outperform conventional

  4. #4 peak

    ash effects can be complex...⏎Whiplash injuries are extremely complex, and profoundly impact many who experience a rear-end collision. Whiplash injuries involve a combination of disruptions in the nervous system, connective tissues, muscles and joints and are often

  5. #5 peak

    consider the size and complexity of an aircraft, you’ll see how involved this process can be. That’s why a Flight Test Instrumentation Engineer works with a team of Engineers, Pilots, and Flight Test Instrumentation Technicians who all have their

  6. #6 peak

    these tars and waxes are no match for an 8000K (7,727°C) plasma arc which, through a combination of heat and UV light, can break them down efficiently. The refinement of the process

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 thinking about lately. I have been thinking about lately. I have been thinking about lately. I have been thinking about lately. I have been thinking about lately. I have been thinking about lately. I have been thinking about lately

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

    built environment around them. Ask students to think of ways that humans influence their environment (streets, houses,<|endoftext|>Food packaging is significant for a few reasons. As a matter

  2. #2 peak

    protein into your diet. This is 13 percent of the daily value for protein, a nutrient your body needs to repair cells and build new ones.⏎Source of Minerals⏎Tamales contain good amounts of iron and phosphorus. Eating a

  3. #3 peak

    even large planes. Vietnam has a developed aviation infrastructure. International air transportation of cargo from Vietnam is carried out by the following airports in the country: Phu Quoc, Noi Baiy, Danang, Katbi, Chulay,

  4. #4 peak

    she is reading into the conversation. Ask questions like: What kind of book is it? What is the setting? What happens? What do you like

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. L0:1621 weakly excites
  2. L0:542 weakly excites
  3. L0:178 weakly suppresses
  4. L0:1744 weakly excites
  5. L0:3014 weakly suppresses
  6. L0:2916 weakly suppresses
  7. L0:500 weakly suppresses
  8. L0:2563 weakly excites
  9. L0:2887 weakly excites
  10. L0:1367 weakly suppresses
  11. L0:2355 weakly excites
  12. L0:986 weakly suppresses
  13. L0:272 weakly excites
  14. L0:1224 weakly suppresses
  15. L0:1301 weakly suppresses
  16. L0:1925 weakly suppresses
  17. L0:437 weakly suppresses
  18. L0:1133 weakly excites
  19. L0:822 weakly suppresses
  20. L0:2753 weakly suppresses
  21. L0:1067 weakly suppresses
  22. L0:170 weakly excites
  23. L0:2652 weakly suppresses
  24. L0:1832 weakly excites

Feeds later layers

  1. L3:1999 weakly excites
  2. L15:799 weakly suppresses
  3. L10:2507 weakly suppresses
  4. L20:2185 weakly suppresses
  5. L17:2345 weakly excites
  6. L20:1756 weakly suppresses
  7. L25:945 weakly suppresses
  8. L27:2307 weakly suppresses
  9. L25:1951 weakly excites
  10. L3:2048 weakly excites
  11. L10:2280 weakly suppresses
  12. L7:549 weakly suppresses
  13. L18:821 weakly excites
  14. L24:1854 weakly suppresses
  15. L3:2794 weakly excites
  16. L22:886 weakly suppresses
  17. L15:1386 weakly excites prepositions after passive verbs
  18. L7:1672 weakly excites
  19. L15:594 weakly excites
  20. L11:247 weakly suppresses
  21. L22:1213 weakly suppresses
  22. L16:2472 weakly suppresses
  23. L5:1771 weakly excites
  24. L2:1946 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 -1.13 and -1.072 values between -0.94 and -0.873 values between -0.67 and -0.612 values between -0.61 and -0.542 values between -0.54 and -0.485 values between -0.48 and -0.419 values between -0.41 and -0.359 values between -0.35 and -0.2837 values between -0.28 and -0.22103 values between -0.22 and -0.15254 values between -0.15 and -0.09635 values between -0.09 and -0.021434 values between -0.02 and 0.04799 values between 0.04 and 0.11379 values between 0.11 and 0.17132 values between 0.17 and 0.2450 values between 0.24 and 0.3022 values between 0.30 and 0.377 values between 0.37 and 0.433 values between 0.43 and 0.506 values between 0.50 and 0.563 values between 0.56 and 0.633 values between 0.63 and 0.693 values between 0.69 and 0.762 values between 0.76 and 0.821 values between 1.02 and 1.081 values between 1.28 and 1.341 values between 1.41 and 1.47offnegativepositiveoften

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. footsteps
  2. gio
  3. uyền
  4. eso
  5. leştir
  6. 几步
  7. utes
  8. uft
  9. vertime
  10. steps

Suppresses

  1. worthy
  2. HomeController
  3. Nullable
  4. òn
  5. entials
  6. -worthy
  7. uler
  8. Quản
  9. 偏离
  10. İyi