Neuron Explorer

Qwen3-0.6B

commas that lead into a continuation

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 a comma that comes after a finished clause, where the sentence is about to add more, like "...difficult to distinguish" or "...including the stomach." It dips on commas inside a list or title. Turning it up made the model repeat words like "which" and "as."

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

    at least the most common variety of obsidian). However, sometimes obsidian isn’t as readily identifiable as you might think. It can grade into other rock types, making it difficult to distinguish.⏎Perhaps you have found a less common variety that doesn

  2. #2 peak

    innovative drugs have been introduced, from Pfizer's (<|endoftext|>Information identified as archived on the Web is for reference,

  3. #3 peak

    have yielded highly successful results, as revealed by Yang. These triumphs pave the way for its maiden voyages

  4. #4 peak

    refer patients to a surgeon when surgical treatment is needed. Many surgeons diagnose and treat disease through nonsurgical methods, but it is usually in an area of medical specialization in which they are considered experts

  5. #5 peak

    to its Cold War importance, but it is yet underappreciated by contemporary policymakers. NATO faces threats from all directions, so how can the Alliance also pay attention to the North Atlantic? How can it

  6. #6 peak

    technical skills are required only in the IT sector, or in office-setting jobs. But that’s not true. Some level of technical

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: , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , ,

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

    of galaxy clusters; they are among the biggest coherent structures in the Universe (what’s bigger? Superclusters, which are clusters of clusters!). Imagine a collection of, oh, say, a thousand galaxies the size of the Milky Way. The sky

  2. #2 peak

    free()⏎- Automatic variables left on the stack of languages like C or garbage collected in languages like Java.⏎Byers, Simon. Scalable Exploitation of, and Responses to Information Leakage Through Hidden Data in Published Documents, AT&T Research,

  3. #3 peak

    posited answers to this question, whether directly - in the case of Plato and Plotinius - or through story - in the case of Homer.⏎Plotinus, having been a proponent of Platonism, naturally concurs with Plato,

  4. #4 peak

    . Think of a world where you could selectively download your life’s memories- sights, sounds, experiences- onto a Blue-ray to show your grandkids (who, by that time, will have something better than Blue-rays). Imagine being able to

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. L14:35 strongly suppresses
  2. L17:118 excites comma after a clause ending
  3. L19:726 suppresses comma inside large numbers
  4. L18:366 excites comma inside numbers and dates
  5. L16:240 excites comma after 'sometimes' or 'the truth'
  6. L11:154 suppresses commas after citation numbers
  7. L9:77 excites commas in citations and name lists
  8. L4:3 excites commas after citation fields
  9. L17:585 excites
  10. L16:241 excites
  11. L18:93 suppresses
  12. L19:1036 suppresses
  13. L15:630 suppresses
  14. L10:335 excites commas in catalog, date, and list metadata
  15. L5:16 excites
  16. L12:1557 suppresses comma after an introductory phrase
  17. L4:96 excites
  18. L16:1059 excites
  19. L15:286 excites
  20. L19:1363 excites
  21. L3:71 suppresses
  22. L16:1939 excites line break mid-sentence in wrapped text
  23. L15:2227 weakly suppresses
  24. L15:1213 weakly suppresses

Feeds later layers

  1. L23:184 excites
  2. L25:880 excites comma after a cause or effect clause
  3. L22:144 excites
  4. L22:576 suppresses
  5. L21:344 suppresses
  6. L27:741 excites
  7. L23:839 excites comma before a trailing 'with' clause
  8. L27:291 suppresses
  9. L27:905 weakly suppresses
  10. L26:236 weakly suppresses
  11. L24:661 weakly excites
  12. L27:677 weakly excites
  13. L21:776 weakly suppresses
  14. L26:765 weakly suppresses
  15. L21:700 weakly excites comma before a new clause
  16. L26:586 weakly suppresses
  17. L24:572 weakly suppresses
  18. L27:86 weakly excites
  19. L21:113 weakly excites
  20. L23:1645 weakly suppresses
  21. L21:1604 weakly excites comma before a trailing phrase
  22. L25:306 weakly suppresses
  23. L25:468 weakly excites
  24. L25:216 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.

2 values between -16.16 and -14.831 values between -12.17 and -10.843 values between -10.84 and -9.521 values between -8.19 and -6.864 values between -6.86 and -5.535 values between -5.53 and -4.207 values between -4.20 and -2.8814 values between -2.88 and -1.551069 values between -1.55 and -0.222406 values between -0.22 and 1.11129 values between 1.11 and 2.4461 values between 2.44 and 3.7741 values between 3.77 and 5.0936 values between 5.09 and 6.4229 values between 6.42 and 7.7517 values between 7.75 and 9.0812 values between 9.08 and 10.416 values between 10.41 and 11.738 values between 11.73 and 13.065 values between 13.06 and 14.393 values between 14.39 and 15.725 values between 15.72 and 17.053 values between 17.05 and 18.387 values between 18.38 and 19.705 values between 19.70 and 21.035 values between 21.03 and 22.366 values between 22.36 and 23.691 values between 23.69 and 25.025 values between 25.02 and 26.341 values between 26.34 and 27.674 values between 27.67 and 29.002 values between 29.00 and 30.331 values between 30.33 and 31.662 values between 34.31 and 35.642 values between 35.64 and 36.97offnegativepositiveoften

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. which
  2. 以及
  3. 也就是
  4. 尤其是在
  5. leading
  6. 包括
  7. 尤其是
  8. including
  9. especially
  10. 其中包括

Suppresses

  1. 、
  2. ),
  3. 的故事
  4. 的好处
  5. ,
  6. 的不同
  7. ,
  8. aven
  9. 的选择
  10. 的要求