Neuron Explorer

Qwen3-0.6B

blank spot right before a number

A neuron in layer 22 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 the empty spot just before a number, such as the gap in 'less than [blank] 0.01' or 'the level of [blank] 280ppm'. When turned up, the model tended to write numbered lists and bracketed numbers.

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

    The ocean’s breadth and depth remain the most extensive, unexplored habit.⏎Activity 2 – Freshwater Habitats⏎Freshwater Habitats cover less than 0.01 % of the earth’s

  2. #2 peak

    79. He completed construction of the Flavian Amphitheatre, better known as the Colosseum, and opened it with ceremonies lasting more than 100 days.⏎What was Tit

  3. #3 peak

    is now above 400ppm and well above the pre-industrial level of 280ppm.⏎After examining the warming evidence Wadhams identifies a

  4. #4 peak

    contours are from the ground-water modeling report. Water-level elevation contours were digitized from a map at a scale of 1:250,000. The maps were published at a

  5. #5 peak

    for all economic activity. Beyond agriculture, water is critical for power generation, mining, industry and the consumer products we rely on every single day. Like the estimated 3000 gallons of water that it takes to make the typical smartphone.⏎

  6. #6 peak

    their drinking water before consumption as well as their water and sanitation conditions.⏎The researchers found that heavy rainfall events (a 24-hour rainfall total that exceeded the 90th percentile value) were associated with an increased risk of diarrheal

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 value of a table's value for a table with a support table with a support table with a support table... with a support table... for a table that's a support table. The value for the support table's value for the value

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

    , ' get its DataSet. Dim thisDataSet As DataSet = CType(DataGrid1.DataSource, DataSet) ' Get the DataTableCollection. Dim tablesCollection As DataTableCollection = thisDataSet.Tables ' Get a specific table by name. Dim table As DataTable =

  2. #2 peak

    of pressure when the heart is resting between beats.⏎Along with high cholesterol and being overweight or obese, a high blood pressure diagnosis, also known as hypertension (over 130/80), is a major risk factor for heart disease or

  3. #3 peak

    pressure.⏎If your blood pressure reading is high, you will probably be asked to come back for repeat blood pressure checks. If you have three visits with readings over 140/90 mmHG, you will be diagnosed with high blood

  4. #4 peak

    decrease your blood pressure through natural methods.⏎Blood pressure is the amount of force your blood pushes against the walls of your blood vessels. A healthy blood pressure is below 120/80.⏎Systolic blood pressure, the top number

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. L2:24 suppresses
  2. L10:370 suppresses
  3. L20:567 excites
  4. L18:66 suppresses
  5. L10:1032 suppresses
  6. L11:223 excites
  7. L13:44 weakly suppresses inside web addresses and citations
  8. L7:92 weakly excites
  9. L18:115 weakly suppresses
  10. L4:76 weakly excites
  11. L8:328 weakly excites
  12. L17:257 weakly suppresses
  13. L16:354 weakly excites
  14. L12:1120 weakly suppresses
  15. L19:305 weakly excites
  16. L19:119 weakly excites
  17. L9:207 weakly suppresses
  18. L15:184 weakly excites
  19. L20:357 weakly excites
  20. L20:253 weakly excites
  21. L10:155 weakly excites
  22. L19:615 weakly excites
  23. L0:1833 weakly excites
  24. L10:873 weakly excites

Feeds later layers

  1. L26:461 excites 'Sch' in surnames
  2. L27:1779 weakly excites
  3. L24:405 weakly excites
  4. L24:1992 weakly suppresses
  5. L26:787 weakly excites
  6. L25:542 weakly excites
  7. L26:1072 weakly suppresses
  8. L27:615 weakly excites
  9. L27:967 weakly excites
  10. L25:386 weakly excites
  11. L23:1928 weakly excites
  12. L23:313 weakly suppresses
  13. L23:758 weakly excites
  14. L24:862 weakly excites
  15. L26:1999 weakly suppresses
  16. L26:348 weakly excites
  17. L26:822 weakly suppresses
  18. L24:1188 weakly suppresses
  19. L23:662 weakly excites
  20. L27:1899 weakly suppresses
  21. L23:553 weakly excites
  22. L23:559 weakly suppresses
  23. L25:746 weakly excites
  24. L24:3027 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 -7.35 and -7.041 values between -6.09 and -5.781 values between -5.46 and -5.141 values between -4.20 and -3.881 values between -3.25 and -2.943 values between -2.94 and -2.622 values between -2.62 and -2.312 values between -2.31 and -1.995 values between -1.99 and -1.683 values between -1.68 and -1.3612 values between -1.36 and -1.0582 values between -1.05 and -0.73355 values between -0.73 and -0.411161 values between -0.41 and -0.101717 values between -0.10 and 0.22438 values between 0.22 and 0.5381 values between 0.53 and 0.8522 values between 0.85 and 1.168 values between 1.16 and 1.484 values between 1.48 and 1.792 values between 1.79 and 2.111 values between 2.42 and 2.741 values between 2.74 and 3.051 values between 3.37 and 3.691 values between 4.00 and 4.321 values between 4.63 and 4.951 values between 4.95 and 5.26offnegativepositiveoften

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. {'
  2. {'
  3. {⏎
  4. { ⏎
  5. {"
  6. ({'
  7. [{'
  8. {⏎⏎
  9. ,{⏎
  10. { ⏎

Suppresses

  1. ustering
  2. ollah
  3. omain
  4. Derby
  5. aise
  6. etable
  7. ieren
  8. دار
  9. igue
  10. lename