Neuron Explorer

Qwen3-0.6B

the word 'still'

A neuron in layer 20 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 'still', as in 'still reflects' or 'still survive today', meaning 'continuing to be'. When it was turned up, the model kept inserting the word 'still' over and over, so the link is clear.

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

    , faculty, postdoctoral scholars, administrators, organizations) served.⏎John Locke ‘s work Some Thoughts Regarding Education was written in 1693 and still reflects conventional education priorities in the Western globe. Every youngster with disability will be identified

  2. #2 peak

    Canada. The work done by the American Bureau was monumental, well informed and incorporated the most advanced scholarship available at the time. In many respects, the information is still useful today, although prudence should be exercised and the reader should consult some of

  3. #3 peak

    is so rare for a lot of people — made him someone who became a touchstone for folks' lives. And that's why they responded to him, and still do.⏎"Keep Ya Head Up" [is] a song that is really

  4. #4 peak

    land so important?⏎- What are the strengths of the Heartland Theory?⏎- What are the weaknesses of the Heartland Theory?⏎- Does the Heartland Theory still apply today?⏎- Key points to remember about Mackinder’s Heartland Theory⏎

  5. #5 peak

    trim- a post he held until 1860. Charles Lanyon firmly made his mark on Victorian Belfast’s urban landscape and many of his works still survive today. Some of his notable works include Queen’s College (184

  6. #6 peak

    80 churches it destroyed was Saint Paul’s Cathedral. The Gothic church was rebuilt by Sir Christopher Wren in Renaissance style and it received its famous dome, which still is, despite fierce competition from the London Eye, one of the most popular viewpoints

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: what if the people who have the most people in the wrong of their people have the most people in the wrong of their wrong ones, and the people who have the most people in the wrong of their wrong ones have the most people in 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

    )o/a|⏎|Beam:||33 ft (10.06 m)|⏎|Draught:||11.5 ft (3.51 m)|⏎|Speed:||16 knots (29.

  2. #2 peak

    Highest point||top of flag staff|⏎|- elevation||15.75 ft (4.8 m) over stone base|⏎July 3, 19

  3. #3 peak

    and had seven locks, each 30 m (98 ft) long, 6 m (20 ft) wide and 1.5 m (4.9 ft) deep. It officially opened in 1825

  4. #4 peak

    an asteroid about 5-10 meters (15-30 feet) in diameter exploding in the air between 15 and 20 km (nine to 12 miles) above sea level. Nobody was injured as a

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. L17:2306 excites
  2. L16:2352 suppresses
  3. L19:2487 excites
  4. L17:2393 suppresses
  5. L16:1470 excites
  6. L16:2301 suppresses
  7. L17:709 excites
  8. L13:2541 excites
  9. L13:2890 excites "still" or "as long as" meaning continuing
  10. L19:2254 suppresses
  11. L10:1848 suppresses
  12. L16:2123 excites
  13. L17:2618 suppresses
  14. L14:2829 weakly excites
  15. L5:2053 weakly suppresses
  16. L18:3044 weakly suppresses
  17. L13:2153 weakly excites
  18. L9:1699 weakly suppresses
  19. L11:2401 weakly excites
  20. L19:1690 weakly excites
  21. L14:1302 weakly excites
  22. L12:2351 weakly excites
  23. L13:1909 weakly suppresses
  24. L8:2072 weakly suppresses

Feeds later layers

  1. L23:1638 strongly excites
  2. L22:2054 suppresses
  3. L26:2595 excites
  4. L27:1280 excites
  5. L22:1795 suppresses
  6. L24:2256 excites
  7. L25:2237 excites
  8. L21:724 weakly excites
  9. L23:2859 weakly excites
  10. L22:1220 weakly excites
  11. L27:2346 weakly excites
  12. L23:525 weakly excites
  13. L22:891 weakly excites
  14. L23:988 weakly suppresses
  15. L21:1772 weakly excites
  16. L26:1243 weakly excites
  17. L22:1432 weakly excites
  18. L22:2508 weakly excites
  19. L22:1774 weakly excites
  20. L24:2111 weakly suppresses
  21. L23:2817 weakly excites the word "no" in "no longer"
  22. L24:1746 weakly excites
  23. L22:2528 weakly excites
  24. L21:2295 weakly 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 -10.20 and -9.131 values between -5.91 and -4.841 values between -3.77 and -2.692 values between -2.69 and -1.62382 values between -1.62 and -0.553416 values between -0.55 and 0.5371 values between 0.53 and 1.6012 values between 1.60 and 2.677 values between 2.67 and 3.753 values between 3.75 and 4.821 values between 4.82 and 5.893 values between 5.89 and 6.971 values between 6.97 and 8.041 values between 8.04 and 9.111 values between 11.26 and 12.331 values between 12.33 and 13.401 values between 17.70 and 18.771 values between 19.84 and 20.921 values between 27.35 and 28.431 values between 31.65 and 32.72offnegativepositiveoften

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. 仍然
  2. 依然
  3. still
  4. 仍是
  5. still
  6. 仍在
  7. 仍然是
  8. 仍
  9. 依旧
  10. 仍有

Suppresses

  1. tab
  2. 已久
  3. ComVisible
  4. inks
  5. аниз
  6. 注
  7. сть
  8. づ
  9. atted
  10. buổi