Neuron Explorer

Qwen3-0.6B

bracket or parenthesis inside a quote

A neuron in layer 15 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 at a point in a quotation where a word has been added or changed, such as a bracketed insertion like '[]us]' or '[sic]', or an opening bracket in a phrase. It is mixed, though, and it also fires strongly on brackets that open a file size or a product note. Steering gave no clear theme.

Written by Claude from this neuron's activations and steering outputs. Confidence: low. Think it is wrong? Flag it (opens an email to the maker with the neuron already filled in).

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

    is carrying out “systematic genocide” in Gaza. South Africa’s ruling party, the African National Congress, asserts that Israel’s actions in Gaza “remind [us] of the atrocities of Nazi Germany.” Palestinian Authority

  2. #2 peak

    many chemotherapy agents are associated with melanonychia, Meuhrcke’s lines, and Mees’ lines, and “it is not uncommon to have one [type of] nail change." But, "having three changes at the same time

  3. #3 peak

    the test proved efficient in the trial phase.⏎“You have a test that is extremely rapid, it seems to be very reliable, and it performs better than PCR, [which] in most cases has between 78 and 90% accuracy

  4. #4 peak

    odge, wrote a little book titled: What is Darwinism? He answered the question himself: It is Atheism!⏎Before long, especially in the South and (as the country expanded) in the West, evolution was taken to be the equivalent

  5. #5 peak

    1-day siege that included several gun battles with federal officers. Means died in 2012 at age 72.⏎“He was the only person in (a) leadership position in Wounded Knee who never left Wounded Knee,

  6. #6 peak

    , stupidity and anger,” said Mangku Puseh Juga, one of 36 priests on hand to sanctify the temple.⏎“The main goal is (to) balance the macro world of nature and the micro world within each human.”⏎

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: could be a thought. So, need to read and restate. Original text. Original text. 1500 words. Repetition. Need to be original. Only copy. 1800 words. Original text. Re

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 fact, you can even write about the kinds of weapons that were used during the war. Consider doing a research on an important personality,<|endoftext|>Download Full Text (640 KB)⏎Decision-making problems in water resources are often associated with multiple

  2. #2 peak

    LPG is a blanket term for two types of natural gas (Propane and Butane) and is a natural by-product of gas<|endoftext|>DataTableCollection.Item Property (String)⏎Gets the DataTable object with the specified name.⏎Assembly: System.Data (

  3. #3 peak

    Based on your location, we recommend that you select: .⏎You can also select a web site from the following list:⏎Select the China site (in Chinese or English) for best site performance. Other MathWorks country sites are

  4. #4 peak

    events and offers. Based on your location, we recommend that you select: .⏎You can also select a web site from the following list:⏎Select the China site (in Chinese or English) for best site performance. Other MathWorks country sites are

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. L10:66 suppresses opening parenthesis after a citation or term
  2. L14:602 suppresses
  3. L11:50 excites
  4. L8:354 excites opening parenthesis before a citation year
  5. L10:1459 suppresses
  6. L4:64 excites
  7. L9:339 excites
  8. L9:2259 suppresses
  9. L9:1421 weakly suppresses
  10. L9:1496 weakly excites
  11. L8:2743 weakly excites
  12. L2:894 weakly excites
  13. L5:330 weakly excites opening parenthesis after a noun
  14. L14:1276 weakly excites
  15. L12:2803 weakly suppresses
  16. L7:732 weakly excites
  17. L12:1443 weakly excites opening parenthesis after a name or term
  18. L7:880 weakly excites
  19. L6:2786 weakly excites
  20. L7:100 weakly suppresses
  21. L12:1557 weakly suppresses comma after an introductory phrase
  22. L10:1244 weakly suppresses
  23. L2:137 weakly excites
  24. L10:2872 weakly excites

Feeds later layers

  1. L19:1119 suppresses
  2. L17:2567 suppresses
  3. L18:2396 suppresses
  4. L17:2234 weakly suppresses
  5. L21:2385 weakly excites
  6. L16:918 weakly suppresses
  7. L16:59 weakly suppresses
  8. L16:1315 weakly suppresses word or phrase being named or defined
  9. L18:1804 weakly suppresses
  10. L25:472 weakly excites
  11. L17:584 weakly suppresses
  12. L17:1171 weakly suppresses
  13. L16:2487 weakly suppresses
  14. L18:1271 weakly suppresses
  15. L18:1481 weakly suppresses
  16. L18:1068 weakly suppresses
  17. L16:1448 weakly suppresses
  18. L16:1053 weakly suppresses
  19. L18:1650 weakly suppresses
  20. L18:1988 weakly suppresses
  21. L17:849 weakly suppresses
  22. L16:1752 weakly excites
  23. L24:1627 weakly suppresses
  24. L18:1491 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 -3.29 and -3.141 values between -2.35 and -2.201 values between -2.04 and -1.881 values between -1.88 and -1.733 values between -1.73 and -1.571 values between -1.57 and -1.412 values between -1.41 and -1.262 values between -1.26 and -1.102 values between -1.10 and -0.953 values between -0.95 and -0.793 values between -0.79 and -0.6317 values between -0.63 and -0.48138 values between -0.48 and -0.32574 values between -0.32 and -0.161365 values between -0.16 and -0.011237 values between -0.01 and 0.15412 values between 0.15 and 0.3195 values between 0.31 and 0.4621 values between 0.46 and 0.6212 values between 0.62 and 0.783 values between 0.78 and 0.934 values between 0.93 and 1.092 values between 1.09 and 1.242 values between 1.24 and 1.401 values between 1.40 and 1.561 values between 1.56 and 1.711 values between 1.87 and 2.031 values between 2.34 and 2.502 values between 2.81 and 2.97offnegativepositiveoften

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. ollywood
  2. blame
  3. 余万元
  4. treffen
  5. 万余元
  6. 有信心
  7. ivities
  8. ://{
  9. ithe
  10. omid

Suppresses

  1. 例外
  2. IEnumerable
  3. icer
  4. itation
  5. description
  6. (description
  7. 见
  8. scription
  9. implementation
  10. ulton