Neuron Explorer

Qwen3-0.6B

location or target word before a noun phrase

A neuron in layer 16 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 on words that set up a place or target, like 'on' in 'points on the sides', 'in' in 'a position in an array', or 'which' before 'bucket'. Turning it up sent the model toward writing about books and files.

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

    . These records are then stored in a number of buckets. We then take the records from the second table and apply the same hash function. This will tell us which bucket to look in. Is there

  2. #2 peak

    to save time.⏎A squad in combat takes a casualty. The medic pulls him behind cover and sees there is blood soaking through his pants but doesn’t know where on the<|endoftext|>In A Theory of the Drone, French philosopher

  3. #3 peak

    oroku" and to pronounce the numbered date that they land on. For example, one students rolls a 3 from Start, they move three spaces and say, "3rd".If a student can't remember the pronunciation

  4. #4 peak

    P inside an equilateral triangle. Draw PA, PB and PC⏎from P perpendicular to the sides of the triangle where A, B and C⏎are points on the sides. Prove that PA + PB + PC is a constant.⏎Can

  5. #5 peak

    position asserted at a trial--the document is often called a brief. It typically consists of a statement of the facts relevant to the case and arguments supported by references to legal authority (statutes, regulations or earlier court decisions). Many briefs are

  6. #6 peak

    have a PhD in English, since your research and writing skills can serve you well. Grant writers need a firm understanding of their field and how to identify the most relevant donors. They also need excellent persuasive writing skills, which includes knowing their audience and

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: I think I have been thinking about the question: "What is the answer to your answer? [1] for each person who the answer to your answer: [1] is? [1] for each person who the answer to your

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

    Administrative Tools Internet Information Services (IIS) Manager. You can access a website via an IP address by expanding the Sites pane of IIS and selecting the website you want to access. You

  2. #2 peak

    1⏎We will use a loop to do that.⏎The values are stored in sequential order in an array and we can access these elements or values by using the index. The index starts from 0 and ends at length of the array -

  3. #3 peak

    -statistical language, with a visual analysis of the means. Then, use Excel to create formulas for a Multiple Means (Fishers) comparison for the two means that differ the most. Show the Excel formulas that are used to do the comparison,

  4. #4 peak

    .⏎And apart from this if you forgot to check this option when recalling the mail, you can use the tracking option. Open the transferred folder and select the mail you recalled. Also, click on the tracing button in the

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. L8:1985 excites
  2. L12:1894 excites
  3. L10:193 excites
  4. L12:2937 excites
  5. L14:2690 excites
  6. L11:3012 excites
  7. L10:1916 excites
  8. L11:952 suppresses
  9. L15:2049 weakly suppresses
  10. L15:2351 weakly excites
  11. L14:2217 weakly suppresses
  12. L14:2888 weakly suppresses
  13. L12:2027 weakly excites
  14. L12:339 weakly excites
  15. L13:2777 weakly suppresses
  16. L15:1355 weakly excites
  17. L13:1915 weakly suppresses
  18. L10:1534 weakly suppresses
  19. L11:2668 weakly excites
  20. L3:1991 weakly suppresses
  21. L12:2353 weakly excites
  22. L12:2938 weakly excites
  23. L10:1689 weakly suppresses
  24. L5:1324 weakly suppresses

Feeds later layers

  1. L18:2235 excites
  2. L17:1258 excites
  3. L17:2022 excites
  4. L19:2128 excites
  5. L20:1038 excites
  6. L17:1782 excites
  7. L17:978 excites
  8. L17:2267 weakly excites
  9. L17:2235 weakly excites
  10. L17:1287 weakly excites
  11. L18:2650 weakly excites
  12. L20:1910 weakly excites
  13. L17:2145 weakly excites
  14. L22:150 weakly suppresses opening parenthesis before an acronym
  15. L17:2633 weakly excites
  16. L24:1541 weakly excites
  17. L17:2855 weakly excites
  18. L20:2879 weakly excites
  19. L18:2146 weakly excites
  20. L18:2899 weakly excites
  21. L24:934 weakly suppresses
  22. L17:1703 weakly excites weak, mixed: 'on' and 'the' after publishing verbs
  23. L17:827 weakly excites
  24. L18:1099 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 -1.58 and -1.451 values between -1.45 and -1.321 values between -1.19 and -1.061 values between -0.93 and -0.803 values between -0.80 and -0.6810 values between -0.68 and -0.5522 values between -0.55 and -0.4267 values between -0.42 and -0.29241 values between -0.29 and -0.16909 values between -0.16 and -0.031812 values between -0.03 and 0.10576 values between 0.10 and 0.23182 values between 0.23 and 0.3648 values between 0.36 and 0.4911 values between 0.49 and 0.6210 values between 0.62 and 0.755 values between 0.75 and 0.871 values between 1.00 and 1.132 values between 1.26 and 1.392 values between 1.39 and 1.521 values between 1.91 and 2.041 values between 3.20 and 3.331 values between 3.46 and 3.59offnegativepositiveoften

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. existing
  2. existing
  3. -existing
  4. (existing
  5. _existing
  6. rè
  7. Existing
  8. الموجودة
  9. Existing
  10. ائن

Suppresses

  1. runApp
  2. gett
  3. cone
  4. щит
  5. luet
  6. swer
  7. iore
  8. Suicide
  9. concussion
  10. serialVersionUID