Neuron Explorer

Qwen3-0.6B

"late" in dated phrases like "in the late 1980s"

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 lights up on the word "late" when it comes right after "in the" and before a decade or century, as in "in the late 1800s." When it was turned up, the model's text drifted toward unrelated rambling, so the steering did not show a clear theme.

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

    ash-spewing hell vent whose name only Icelanders could pronounce—the B.C. volcanologist might have been best equipped to rationalize her frustration. Back in the late 1980s, Hickson had helped spearhead efforts to create

  2. #2 peak

    ".⏎Famed British author Norman Davies noted that American forces fought with distinction on two fronts, but that it was our home factories that turned the tide. In the late 1930's, American industry produced consumer goods at a dog's

  3. #3 peak

    permitted lamps to burn whale-oil and, later, kerosene—today, these are the primary kinds of antique lamps most favored by collectors. In the late 18th century, Swiss chemist Aime Argand invented the first

  4. #4 peak

    Federal authorities for her Southern nationalism. A younger brother, Samuel, was a major in the Georgia infantry and spent time as a Union prisoner of war.⏎In the late 1840s, the Levy family left Charleston for

  5. #5 peak

    explanation. As the last remaining fabric from a series of four, Oxford cloth’s story is one of inexplicable survival. In the late nineteenth century a Scottish Mill designed and produced four cloths, one for each of

  6. #6 peak

    political discussion and debate in digital space having on real life politics? How is the Internet changing actual politics?⏎When the Internet began arriving in the Middle East in the late 1990s, some had high expectations that it would usher in

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, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the,

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

    the uprising⏎of the British Empire. But we have no right to this position of special⏎and honorable responsibility unless we are determined to use it to the⏎full, not for our own

  2. #2 peak

    Should the money follow the people, who have flocked for decades to the Sun Belt, where homes are cheap and jobs are plentiful but climate risks mount by the day? Or, following the logic of Biden’s own order, should funds be directed

  3. #3 peak

    a goat wants to go somewhere, and you don’t have a good hold on her, she’ll drag you along for the ride. She’s usually heading for the nearest fruit<|endoftext|>In a few days a carrier rocket

  4. #4 peak

    each other to no avail. Each person’s actions were canceled out by the others’ actions. The sub just sat there.⏎Finally, from the back of the room, Carpenter’s voice suggested, “Why don’t you guys go

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. L2:205 excites
  2. L3:299 weakly excites
  3. L3:344 weakly excites
  4. L3:2904 weakly excites
  5. L0:2693 weakly suppresses
  6. L0:1061 weakly suppresses
  7. L4:1984 weakly suppresses
  8. L4:2904 weakly excites
  9. L3:2998 weakly suppresses
  10. L2:2703 weakly suppresses
  11. L1:2239 weakly excites
  12. L1:2056 weakly suppresses
  13. L2:1439 weakly excites
  14. L4:455 weakly suppresses
  15. L4:12 weakly suppresses
  16. L3:2698 weakly suppresses
  17. L2:1880 weakly suppresses
  18. L0:1197 weakly excites
  19. L1:491 weakly excites
  20. L1:1244 weakly suppresses "of" after "board" or "off"
  21. L1:507 weakly excites
  22. L1:408 weakly excites
  23. L1:86 weakly suppresses
  24. L2:2388 weakly suppresses

Feeds later layers

  1. L6:2164 strongly excites second half of multi-word proper names
  2. L8:849 suppresses
  3. L10:457 suppresses
  4. L7:26 suppresses
  5. L6:2283 suppresses
  6. L9:716 excites
  7. L9:1607 suppresses
  8. L10:1316 excites
  9. L9:929 excites 'each' in 'each other'
  10. L14:1403 excites
  11. L16:662 suppresses
  12. L16:656 suppresses
  13. L10:217 suppresses
  14. L10:952 suppresses
  15. L6:1463 excites
  16. L15:1601 suppresses
  17. L8:98 excites
  18. L9:845 excites
  19. L7:101 excites
  20. L13:1550 suppresses
  21. L6:1354 excites
  22. L12:576 excites 'of' in fixed phrases like table of contents
  23. L7:111 excites
  24. L13:1513 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 -1.78 and -1.661 values between -1.66 and -1.541 values between -1.05 and -0.932 values between -0.93 and -0.814 values between -0.69 and -0.579 values between -0.57 and -0.4547 values between -0.45 and -0.33190 values between -0.33 and -0.21903 values between -0.21 and -0.081687 values between -0.08 and 0.04777 values between 0.04 and 0.16203 values between 0.16 and 0.2840 values between 0.28 and 0.4010 values between 0.40 and 0.528 values between 0.52 and 0.644 values between 0.64 and 0.764 values between 0.76 and 0.884 values between 0.88 and 1.013 values between 1.01 and 1.131 values between 1.25 and 1.371 values between 1.49 and 1.612 values between 1.61 and 1.731 values between 1.73 and 1.851 values between 1.85 and 1.971 values between 1.97 and 2.091 values between 2.34 and 2.462 values between 2.94 and 3.06offnegativepositiveoften

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. UMP
  2. ative
  3. 下去
  4. tastes
  5. umps
  6. ものが
  7. hon
  8. �
  9. ��
  10. ствие

Suppresses

  1. (Mat
  2. FixedSize
  3. -CS
  4. (Role
  5. >No
  6. pagesize
  7. arness
  8. (Job
  9. tyres
  10. -lat