Neuron Explorer

Qwen3-0.6B

unclear: mid-phrase words like 'difference' and 'liminal'

A neuron in layer 6 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

The strongest moments are scattered: words like 'difference', 'liminal' and 'Nile' inside longer phrases. I could not find a clear pattern, so this one is hard to describe. Turning it up gave text on many unrelated topics.

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

    helping to settle the conflict in Northern Ireland.⏎Freedland writes, the role of the Irish-American diaspora in the Northern Ireland peace process, made a huge difference.⏎“Long before the first IRA ceasefire of 1994, several

  2. #2 peak

    schooling, open parks, and playgrounds will return. But, we do have a sense of what we can—and must—do in this moment, this liminal time.<|endoftext|>When we were in kindergarten and playing sports and our team lost,

  3. #3 peak

    loyal retainer with high hopes for the young queen, playing a pivotal role in the Alexandrine Civil War, but in the aftermath of the Battle of the Nile, Cleopatra and her newfound Roman ally Julius Caesar joined with the Order and

  4. #4 peak

    blur, the deliberate imprecision and smoky dissolution of the formality of the painting is indicative of the manifestation of abstraction and expressionism. Owning this liminal space of amalgamating tradition and contemporary, Akhlaq went on to

  5. #5 peak

    birds ranged from a Red-Throated Hummingbird to an American Goldfinch, a Grey Goose and so much more! Once they reached Bate Island

  6. #6 peak

    for recreational use⏎- Understand how marijuana works in the brain⏎At the completion of this program, participates will know:⏎- The difference between two compounds in the marijuana plant, THC and CBD, and how they affect

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: 10000000000000000000000000000000000000000000000

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

    energy needs. For the majority of off-grid operations, this means a choice between oil, LPG (liquefied petroleum gas) or LNG

  2. #2 peak

    with the exception of pieces that are being conserved.⏎The original marble reliefs at the Folger have encountered various conservation issues throughout the years, particularly in the form of erosion due to acid rain.

  3. #3 peak

    Latin America.⏎City Fragility and Community Resilience⏎In South Asia, Karachi, Pakistan, is an example of a city that exhibits fragility in the form of violence, and resilience in the form of community-led initiatives. In particular areas

  4. #4 peak

    aim for medium heat with these foods. For a gas grill, this means reducing the temperature, and for charcoal, it means building a smaller fire. You

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. L3:2079 weakly excites
  2. L2:1516 weakly suppresses
  3. L5:2391 weakly excites
  4. L4:2421 weakly excites
  5. L5:1975 weakly excites
  6. L3:405 weakly suppresses
  7. L4:1701 weakly excites
  8. L4:552 weakly suppresses
  9. L2:2923 weakly suppresses
  10. L2:1314 weakly excites
  11. L2:1013 weakly suppresses
  12. L4:2507 weakly excites
  13. L4:1862 weakly excites
  14. L5:64 weakly excites
  15. L5:253 weakly suppresses
  16. L3:1799 weakly suppresses
  17. L3:535 weakly excites
  18. L3:1744 weakly excites
  19. L5:1212 weakly excites
  20. L5:1576 weakly excites
  21. L2:969 weakly excites
  22. L3:1887 weakly suppresses
  23. L0:3066 weakly suppresses
  24. L0:482 weakly suppresses

Feeds later layers

  1. L12:1248 suppresses
  2. L24:1851 weakly excites
  3. L22:1824 weakly excites
  4. L13:1949 weakly excites
  5. L10:673 weakly excites
  6. L14:1072 weakly excites
  7. L12:2630 weakly excites
  8. L13:519 weakly suppresses
  9. L15:1575 weakly suppresses
  10. L12:18 weakly suppresses
  11. L23:1273 weakly suppresses
  12. L14:1026 weakly suppresses
  13. L7:2560 weakly excites
  14. L7:1359 weakly suppresses
  15. L26:825 weakly suppresses
  16. L14:2945 weakly excites
  17. L15:1046 weakly suppresses
  18. L8:1039 weakly suppresses
  19. L20:2473 weakly excites
  20. L27:1348 weakly suppresses
  21. L19:1354 weakly excites
  22. L22:1453 weakly excites
  23. L9:2962 weakly excites
  24. L21:3008 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 -1.72 and -1.611 values between -1.61 and -1.501 values between -1.50 and -1.391 values between -1.39 and -1.282 values between -1.28 and -1.173 values between -1.17 and -1.067 values between -1.06 and -0.953 values between -0.95 and -0.843 values between -0.84 and -0.733 values between -0.73 and -0.6210 values between -0.62 and -0.5122 values between -0.51 and -0.3963 values between -0.39 and -0.28196 values between -0.28 and -0.17686 values between -0.17 and -0.061966 values between -0.06 and 0.05580 values between 0.05 and 0.16208 values between 0.16 and 0.2773 values between 0.27 and 0.3831 values between 0.38 and 0.4918 values between 0.49 and 0.6012 values between 0.60 and 0.7110 values between 0.71 and 0.821 values between 0.82 and 0.933 values between 1.04 and 1.151 values between 1.15 and 1.262 values between 1.26 and 1.371 values between 2.59 and 2.70offnegativepositiveoften

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. 其他国家
  2. Indian
  3. 想不到
  4. ing
  5. Indian
  6. contrad
  7. States
  8. 节目
  9. uto
  10. ace

Suppresses

  1. 十二条
  2. lake
  3. 岷
  4. сон
  5. ñe
  6. mó
  7. ---</
  8. 倒在
  9. ieri
  10. ți