Neuron Explorer

Qwen3-0.6B

comma after "however" or "of course"

A neuron in layer 3 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 responds to the comma that comes after a word like "however" or "of course" in the middle of a sentence. Its negative side is for split-up word pieces, such as the middle of "bioproducts" or "understand". Turning it up did not produce a clear theme in the steered text.

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

    full page advertisement in the Trolley pamphlet for a night run to New York - first class - at $1.50. For shorter journeys, though, the trolley was hard to beat.⏎Hartford needs far more than<|endoftext|>

  2. #2 peak

    political thought. Their lives were quite tragic, of course, as Kaczynski murdered several people, and McCandless accidently killed

  3. #3 peak

    Wilson disease, a disease caused by an overload of copper, and certain metastatic cancers. That much is known. Very little, however, is known about how the drug works at the molecular level.A new study led by

  4. #4 peak

    until this point, his paintings were of peasant life in sombre colours. Once settled in Paris, however, his paintings took on new painting styles and brighter colours, and he started painting urban

  5. #5 peak

    our area, this arrives roughly around Thanksgiving. At this point, many plant species significantly slow down their productivity and go dormant.⏎Many vegetable plants do not, however, die and decay away. Rather, they will hold their vitality and nutritional value through

  6. #6 peak

    is also a disappointment, for she finds the conversation of both men and women personal and trivial. She tries to introduce something different. On the way home, however, she is lectured by her husband on the danger of shocking people.⏎As time

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 problem of a single person. The problem of a single person is a subject of the problem of a single person. The problem of a single person is a subject of the problem of a single person. The problem of a single person is

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

    at Western Washington University.⏎The Northwest Advanced Renewables Alliance (NARA) Education group aims to:⏎- Meet the workforce needs of the bio-energy/bioproducts economy⏎- Develop a broad, integrated view of the biofuels problem

  2. #2 peak

    WI) residents. The ultimate goal of this resource is to reduce the occurrence of water-related illness amongst Wisconsin’s private well users.⏎Goals of the Program:⏎Understand current process and identify system needs related to communicating with the public about well water quality

  3. #3 peak

    available for voice and data networks, their transmission speeds (bandwidth), and how they are packaged for commercial use. CS610 Handouts pdf⏎Understand the importance of adapting a Computer Network offering to meet the changing and challenging networking needs

  4. #4 peak

    ). This may be further exacerbated by few harmful chemicals (proteins) such as tumor necrosis factor alpha (TNF-α) and chem

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. L0:1430 excites
  2. L1:1418 excites
  3. L1:298 suppresses
  4. L2:1831 excites
  5. L1:1075 excites
  6. L1:84 excites
  7. L2:1604 suppresses
  8. L0:2643 suppresses
  9. L1:1026 suppresses
  10. L1:2866 excites
  11. L2:2014 suppresses
  12. L1:58 suppresses
  13. L1:1133 excites
  14. L0:859 excites
  15. L0:2305 suppresses
  16. L1:39 weakly suppresses
  17. L0:2538 weakly suppresses
  18. L0:491 weakly excites
  19. L0:1791 weakly suppresses
  20. L2:2128 weakly excites
  21. L0:2317 weakly suppresses
  22. L1:482 weakly suppresses
  23. L0:925 weakly suppresses
  24. L0:2972 weakly excites

Feeds later layers

  1. L20:87 strongly excites
  2. L17:123 excites
  3. L19:276 suppresses
  4. L18:340 excites
  5. L16:484 suppresses
  6. L21:85 suppresses
  7. L19:765 excites
  8. L17:282 excites
  9. L16:379 excites
  10. L16:1609 excites
  11. L4:1067 excites
  12. L19:382 excites
  13. L6:1276 excites
  14. L17:96 weakly suppresses
  15. L6:360 weakly excites
  16. L20:43 weakly excites
  17. L20:418 weakly excites
  18. L15:2803 weakly excites
  19. L20:687 weakly excites
  20. L27:2856 weakly suppresses
  21. L6:2181 weakly excites
  22. L4:47 weakly suppresses
  23. L6:163 weakly suppresses
  24. L20:472 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 -1.09 and -1.041 values between -1.04 and -0.991 values between -0.99 and -0.952 values between -0.95 and -0.903 values between -0.85 and -0.802 values between -0.76 and -0.714 values between -0.71 and -0.664 values between -0.66 and -0.616 values between -0.61 and -0.5713 values between -0.57 and -0.526 values between -0.52 and -0.479 values between -0.47 and -0.4235 values between -0.42 and -0.3848 values between -0.38 and -0.3348 values between -0.33 and -0.2878 values between -0.28 and -0.23111 values between -0.23 and -0.19146 values between -0.19 and -0.14215 values between -0.14 and -0.09367 values between -0.09 and -0.05723 values between -0.05 and 0.00605 values between 0.00 and 0.05338 values between 0.05 and 0.10254 values between 0.10 and 0.15208 values between 0.15 and 0.19196 values between 0.19 and 0.24147 values between 0.24 and 0.29116 values between 0.29 and 0.3380 values between 0.33 and 0.3866 values between 0.38 and 0.4338 values between 0.43 and 0.4818 values between 0.48 and 0.525 values between 0.52 and 0.578 values between 0.57 and 0.622 values between 0.62 and 0.671 values between 0.67 and 0.711 values between 0.71 and 0.762 values between 0.76 and 0.81offnegativepositiveoften

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. 4
  2. itis
  3. (
  4. 是
  5. tab
  6. our
  7. Tab
  8. 内容
  9. Inc
  10. tab

Suppresses

  1. "]."
  2. ##⏎⏎
  3. bnb
  4. blatant
  5. isia
  6. 答卷
  7. топ
  8. bia
  9. ]+\
  10. BorderStyle