Neuron Explorer

Qwen3-0.6B

unclear: words before an explanation

A neuron in layer 20 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on words that set up an explanation or a list, such as 'due' in 'was due to', or an opening bracket before a note. Its negative side fires on the word 'a' before a role name, like 'a lexicographer'. Turning it down made the model produce empty blank-filled templates, but turning it up had no clear effect.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. 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

    frequent intervals becomes.⏎When initial temperature is measured in those products which incorporate frozen ingredients extra care may need to be taken to insure that the temperature reading is of the (average) temperature of the contents of the container and not just the temperature of the

  2. #2 peak

    with fuel. For centuries, it was firewood, but by the early 1900s the use of firewood was increasingly inconvenient.⏎This was due to two major problems. First of all, population growth meant that most mountains around

  3. #3 peak

    not everyone has had their own experience with SAD, many have watched someone else go through it. When talking about seeing someone else go through this, (Male 282) says, “they’re different, distant, standoff

  4. #4 peak

    1___ in the United States and, until the beginning of the 20th century, was used by nearly all American cities. Its structure is similar to that of the ___2___ and national governments, with an elected mayor as chief of

  5. #5 peak

    . For such individuals, concentrations in information systems, information technology, or computer studies generally would be more appropriate.⏎These recommendations for formulating concentrations in computer science are based on "Computing Curricula 2001: Computer Science" by

  6. #6 peak

    by an average of 5.6% per year from 2009-13 with a stable exchange rate and low inflation. This growth was due partly to high international prices for Peru's metals and minerals exports, which account for

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: (____) (____) (____) (____) (____) (____) (____) (____) (____) (____) (____) (____) (____) (____) (____) (____)

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

    best access the data incorporated in printed and electronic dictionaries. This is sometimes referred to as 'metalexicography'.⏎A person devoted to lexicography is called a lexicographer.⏎General lexicography focuses on the design, compilation, use

  2. #2 peak

    us, Species⏎List the taxa in order from largest to smallest.⏎A heterotrophic eukaryote that absorbs nutrients from organic materials in the environment is a.⏎The system

  3. #3 peak

    memory, which does not have a ROM, to bring the loader to the memory once at start up.⏎A microcomputer system with a ROM and a RAM has a monitoring program in its ROM in any case, so the computer can be completely operated

  4. #4 peak

    most commonly affect women and occur in adolescence and young adulthood, though they can occur in anyone, of any gender, at any age.⏎Eating disorders most commonly affect women and occur in adolescence and young adulthood, though they can occur in anyone,

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. L17:2664 strongly excites
  2. L19:2077 strongly excites
  3. L18:2255 strongly suppresses
  4. L15:1538 excites
  5. L16:685 excites
  6. L17:2866 suppresses
  7. L18:2845 excites
  8. L19:2164 suppresses
  9. L16:1199 excites
  10. L19:1193 suppresses
  11. L18:2185 suppresses
  12. L18:394 excites
  13. L17:644 excites
  14. L12:2186 excites
  15. L13:2012 excites
  16. L18:2819 suppresses
  17. L17:623 suppresses
  18. L16:1965 excites
  19. L19:1246 excites
  20. L18:419 weakly suppresses
  21. L14:441 weakly excites
  22. L19:2591 weakly suppresses
  23. L15:2158 weakly suppresses
  24. L17:371 weakly suppresses

Feeds later layers

  1. L25:360 strongly suppresses
  2. L26:234 strongly suppresses
  3. L23:224 strongly suppresses
  4. L26:13 strongly excites
  5. L22:351 suppresses
  6. L27:218 suppresses
  7. L21:416 suppresses
  8. L27:921 suppresses
  9. L26:1589 suppresses
  10. L27:475 suppresses
  11. L27:904 suppresses
  12. L25:300 suppresses
  13. L26:256 suppresses
  14. L25:474 suppresses
  15. L27:1373 suppresses
  16. L26:378 suppresses
  17. L22:161 suppresses
  18. L22:317 excites
  19. L26:1059 suppresses
  20. L24:146 suppresses
  21. L25:51 excites
  22. L25:161 suppresses
  23. L24:260 suppresses
  24. L22:795 excites

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 -21.66 and -20.971 values between -18.20 and -17.511 values between -17.51 and -16.822 values between -15.44 and -14.751 values between -14.75 and -14.061 values between -14.06 and -13.371 values between -13.37 and -12.682 values between -12.68 and -11.995 values between -11.99 and -11.292 values between -11.29 and -10.606 values between -10.60 and -9.912 values between -9.91 and -9.2210 values between -9.22 and -8.536 values between -8.53 and -7.8414 values between -7.84 and -7.159 values between -7.15 and -6.4618 values between -6.46 and -5.7718 values between -5.77 and -5.0838 values between -5.08 and -4.3952 values between -4.39 and -3.7063 values between -3.70 and -3.00102 values between -3.00 and -2.31156 values between -2.31 and -1.62240 values between -1.62 and -0.93474 values between -0.93 and -0.241778 values between -0.24 and 0.45817 values between 0.45 and 1.1484 values between 1.14 and 1.831 values between 1.83 and 2.521 values between 2.52 and 3.211 values between 3.21 and 3.901 values between 5.29 and 5.98offnegativepositiveoften

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. exampleInput
  2. &apos
  3. odox
  4. 듯
  5. &T
  6. eful
  7. -FIRST
  8. xhttp
  9. JsonValue
  10. eyJ

Suppresses

  1. **
  2. ____
  3. __
  4. **
  5. ___
  6. ____
  7. ______
  8. (__
  9. (**
  10. ________