Neuron Explorer

Qwen3-0.6B

'both X and Y' pairing words

A neuron in layer 21 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 responds to words in sentences like 'both ... and', 'both excited and terrified', or 'between 1910 and 1930' where two things are being paired. It is weaker on the number sections of date ranges. When it was turned up, the model produced loose, unfocused text, and turning it down made it repeat 'and'.

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

    by adding English to Urdu translations, Urdu to Roman Urdu transliterations and Urdu to English Translations. This will improve our English to Urdu Dictionary, Urdu to English dictionary, English to Urdu Idioms translation and Urdu to English Idioms translations. Although

  2. #2 peak

    determined to continue flying. Arthur Whitten Brown, born 1886, was, after WW1 determined to develop his navigational skills. Both Alcock and Brown were dedicated to becoming highly proficient aviators.⏎In April 19

  3. #3 peak

    programs from around the world — a vision of worldwide broadcasts from an age when TV was little more than an experimental toy for the rich.⏎But designs like these both excited and terrified people in the radio industry. By 1935, people

  4. #4 peak

    “To combat global warming, we’re understandably going to make our buildings more sustainable. But in that process, we need to think of sustainability as both reducing energy demand and protecting the health of the building occupants.”⏎Waring’s academic background is as

  5. #5 peak

    as well as quality protein, whole grains and low-fat dairy products or affordable plant milks. Regular, physical activity (both weight-bearing for musculoskeletal health benefits and intensive for cardiovascular health benefits) would be part of normal life. Maybe this

  6. #6 peak

    soil moist all the way through. It also controls the growth of weeds. Mulch, therefore, fulfils the dual purpose of keeping the weeds away as well as providing nutrients to the soil.⏎But what exactly is mulch?

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: and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and and

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

    relief. The promise of jobs in northern factories during World War I provided African Americans an opportunity to escape the harsh realities of the South. Between 1910 and 1930, approximately 1.6 million African Americans left

  2. #2 peak

    world.1⏎A 2007 report by the World Bank called attention to the rapid gains made by MENA countries:⏎Between 1980 and 2010, the average years of schooling for females aged

  3. #3 peak

    470,000 people—or 21 percent of the island’s total population—left Puerto Rico for the United States between 1950 and 1960.⏎In what ways did the counterculture represent

  4. #4 peak

    leaders also knew how to obtain it. They were themselves usually soldiers, and knew how to organize large, complex ventures along military lines. Between 1660 and 1727, 87% of the colonial governors were

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. L20:68 strongly excites
  2. L19:979 strongly excites
  3. L16:357 strongly suppresses
  4. L19:612 strongly excites
  5. L20:348 strongly suppresses
  6. L16:1669 strongly suppresses
  7. L15:1542 strongly excites
  8. L17:2776 strongly suppresses
  9. L18:2118 strongly excites
  10. L18:1011 strongly suppresses
  11. L15:2643 strongly excites
  12. L19:257 strongly suppresses
  13. L19:1245 strongly suppresses comma opening a list of parallel items
  14. L16:2500 strongly excites
  15. L19:2069 strongly excites
  16. L19:2729 strongly excites
  17. L16:2279 strongly suppresses
  18. L19:2726 strongly suppresses
  19. L20:2098 strongly excites
  20. L15:1924 strongly suppresses
  21. L13:557 excites
  22. L16:2738 excites
  23. L20:2544 excites
  24. L17:2911 suppresses

Feeds later layers

  1. L24:205 strongly suppresses
  2. L27:289 strongly suppresses
  3. L25:6 strongly excites
  4. L26:65 strongly excites
  5. L26:347 strongly suppresses the 're' in self-reliant and similar
  6. L23:379 strongly suppresses
  7. L22:632 strongly suppresses
  8. L25:1270 strongly excites
  9. L22:791 strongly suppresses
  10. L25:583 suppresses
  11. L26:74 excites
  12. L25:801 suppresses
  13. L26:2252 excites
  14. L27:1902 excites
  15. L27:2212 suppresses
  16. L24:289 suppresses
  17. L26:2056 excites
  18. L27:1937 suppresses
  19. L27:1364 suppresses
  20. L26:307 weakly suppresses
  21. L26:859 weakly suppresses
  22. L25:120 weakly excites
  23. L27:291 weakly suppresses
  24. L26:1868 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.

2 values between -9.51 and -9.241 values between -8.14 and -7.871 values between -7.60 and -7.331 values between -6.78 and -6.511 values between -6.51 and -6.231 values between -5.96 and -5.692 values between -5.69 and -5.412 values between -5.41 and -5.142 values between -5.14 and -4.872 values between -4.87 and -4.601 values between -4.60 and -4.323 values between -4.32 and -4.057 values between -4.05 and -3.785 values between -3.78 and -3.5011 values between -3.50 and -3.239 values between -3.23 and -2.9611 values between -2.96 and -2.6918 values between -2.69 and -2.4126 values between -2.41 and -2.1422 values between -2.14 and -1.8733 values between -1.87 and -1.5951 values between -1.59 and -1.3262 values between -1.32 and -1.0567 values between -1.05 and -0.78138 values between -0.78 and -0.50203 values between -0.50 and -0.23608 values between -0.23 and 0.041084 values between 0.04 and 0.32894 values between 0.32 and 0.59465 values between 0.59 and 0.86146 values between 0.86 and 1.1429 values between 1.14 and 1.41offnegativepositiveoften

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. /or
  2. acles
  3. phans
  4. withstanding
  5. ˌ
  6. srv
  7. ære
  8. utr
  9. 蔼
  10. 냄

Suppresses

  1. and
  2. 和
  3. 和个人
  4. và
  5. 和技术
  6. และ
  7. 和他的
  8. 和发展
  9. and
  10. 和社会