Neuron Explorer

Qwen3-0.6B

unclear: sentence and clause breaks in essays

A neuron in layer 22 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 weakly at scattered points in explanatory writing, such as the start of a sentence or the spot after a comma or full stop. There is no clear topic, and turning it up or down produced text with no obvious link to it.

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

    Chemical Activity of Copper:⏎The chemical activity of copper imparts unique taste and aromatic characteristics to beverages.⏎Despite their

  2. #2 peak

    reasons why copper stills are highly valued:⏎Excellent Thermal Conductivity of Copper:⏎Copper possesses outstanding thermal conductivity, surpassing stainless steel by 24 times. This is particularly crucial in distillation, where even heating of the mash ensures more

  3. #3 peak

    the potential to strike numerous targets concurrently, from considerable distances (hundreds of kilometers), with a relatively low cost for each

  4. #4 peak

    ⏎Mary Oliver, a renowned American poet, has captivated readers with her profound exploration of nature and the human experience. Her poems often evoke a sense of wonder and contemplation, inviting readers to reflect on their

  5. #5 peak

    dilemmas have surfaced. For example, should people be tested to determine whether they might later develop a disease if

  6. #6 peak

    bodies, and pigment granules. Understanding these components will not only deepen your knowledge of hair but also help you identify them and comprehend their significance. So, let's dive in!⏎The cuticle is the outermost layer of the hair

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: how about using a different way of how people and people (main and main main and main main main and main main main main main and main main main main main main and main main main main main main main and some different way of using (

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

    not protect against COVID-19, and the COVID-19 vaccine will not protect against flu. Moreover, getting a flu shot<|endoftext|>Housing Garden TortoisesHERMANNS, SPUR-THIGHS, MARGINATED⏎

  2. #2 peak

    to facilitate the shedding of the uterine lining. Uterine polyps can vary in size, with the smallest being only a few millimeters wide. The growths mostly occur in women who are

  3. #3 peak

    , and mood swings can significantly affect your life and make it almost impossible for you to enjoy yourself fully. In some cases, menopause symptoms may be severe to the point of preventing you from carrying out your daily activities. Some of the symptoms you

  4. #4 peak

    required to have one given name and one family name, but can have as many additional given names and middle names as they like. A middle name is usually a recognised surname and not a patronym

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. L20:748 strongly excites
  2. L16:704 strongly excites
  3. L15:2567 strongly suppresses
  4. L18:232 suppresses
  5. L10:1050 suppresses end of document marker token
  6. L17:581 suppresses
  7. L18:454 excites
  8. L10:58 excites
  9. L18:1141 excites unclear: mid-sentence words before a noun
  10. L9:1378 excites
  11. L3:112 excites
  12. L10:1523 suppresses
  13. L14:160 excites
  14. L10:557 excites
  15. L11:533 suppresses
  16. L21:677 excites
  17. L7:541 excites
  18. L14:1266 excites
  19. L11:308 excites
  20. L9:51 excites
  21. L19:117 excites
  22. L19:967 suppresses
  23. L19:686 excites
  24. L9:260 suppresses

Feeds later layers

  1. L27:1096 excites
  2. L24:282 excites the 'conc' word fragment
  3. L26:2773 excites
  4. L27:1299 excites
  5. L27:196 excites
  6. L26:1456 excites
  7. L25:2469 excites
  8. L26:579 excites
  9. L25:308 excites
  10. L25:2204 suppresses
  11. L27:322 weakly excites
  12. L25:456 weakly suppresses
  13. L23:273 weakly suppresses
  14. L24:602 weakly excites
  15. L27:172 weakly suppresses
  16. L23:1093 weakly suppresses
  17. L27:532 weakly suppresses
  18. L27:1777 weakly excites
  19. L27:402 weakly suppresses
  20. L25:309 weakly excites
  21. L25:420 weakly suppresses
  22. L24:19 weakly excites
  23. L27:2494 weakly suppresses
  24. L24:2862 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.

1 values between -3.78 and -3.551 values between -2.84 and -2.611 values between -2.61 and -2.373 values between -2.13 and -1.903 values between -1.90 and -1.661 values between -1.66 and -1.433 values between -1.43 and -1.199 values between -1.19 and -0.9647 values between -0.96 and -0.72106 values between -0.72 and -0.48382 values between -0.48 and -0.251144 values between -0.25 and -0.011543 values between -0.01 and 0.22451 values between 0.22 and 0.46128 values between 0.46 and 0.7037 values between 0.70 and 0.9314 values between 0.93 and 1.1712 values between 1.17 and 1.404 values between 1.40 and 1.643 values between 1.64 and 1.873 values between 1.87 and 2.114 values between 2.11 and 2.352 values between 2.35 and 2.581 values between 2.58 and 2.821 values between 2.82 and 3.051 values between 4.00 and 4.231 values between 4.70 and 4.941 values between 5.17 and 5.411 values between 5.41 and 5.65offnegativepositiveoften

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. forthcoming
  2. contemplate
  3. paramount
  4. burgeoning
  5. bestowed
  6. amassed
  7. embarked
  8. elucid
  9. adorned
  10. repertoire

Suppresses

  1. 怎么做
  2. 很漂亮
  3. loud
  4. ichtig
  5. úng
  6. ách
  7. пло
  8. 写的
  9. 没事
  10. 非常好