Neuron Explorer

Qwen3-0.6B

large number scale words

A neuron in layer 4 of Qwen3-0.6B. Its strongest snippets are fairly consistent: several of them peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron responds to big number words like 'billion', 'trillion' and 'bn' that come right after a figure. When it was turned up, the model wrote out large numbers and population counts.

Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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

    orbiting satellites collecting solar energy and blasting it down to receiving stations in high powered laser beams. Safe free energy forever .... all we need are a few hundred thousand billions to set it up.<|endoftext|>Built on our successful pilot work in two Bay Area

  2. #2 peak

    internet becomes integrated into virtually every aspect of life, keeping its infrastructure secure will be ever more essential. Spending on cybersecurity globally is expected to surpass $200bn in 2

  3. #3 peak

    The development of hydraulic fracturing (‘fracking’) has raised US gas production dramatically, and helped double the estimated global gas resource to around 800 trillion cubic meters (tcm), or 250 years’ consumption at current

  4. #4 peak

    of processors. The country had crossed the petaflop computer range and needs exascale computers, which are machines capable of performing 1000 trillion operations a second."⏎IIT Delhi will provide the infrastructure for the Exascale

  5. #5 peak

    in about a cup of DNA.⏎There is a lot of digital information in the world about three zettabytes' worth (that's 3000 billion billion bytes) and the constant influx of new digital content poses a real challenge for

  6. #6 peak

    the power system by 2030, as much as 15 trillion litres of water could be saved. That's the equivalent to the annual consumption of

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 am doing, something that is I am doing, and so I am (something) This is the same as I am (something) This is the same as I am This is the same as I am This is the

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 the neuron network of a human brain, with ultra-high resolution and in three dimensions. The coalition now grows to a network involves over 1,000 researchers from Singapore, Australia, China, Japan, Korea and Taiwan. The team

  2. #2 peak

    as "the Monster" to the scientists who conjured it – to explore the mysterious subatomic realm. Fifty years later, more than 1,000 people gathered at SLAC National Accelerator Laboratory on Friday and Saturday to celebrate the

  3. #3 peak

    dust generation inside pebble-bed reactors.⏎MELCOR is a fully integrated, engineering-level computer code that models the progression of severe accidents in light water reactor nuclear power plants. MELCOR is being developed at Sandia National Laboratories (

  4. #4 peak

    023/B:BIOC.0000048451.07820.9c⏎- 1.1k Downloads⏎Recent massive imports of slider turtles, (Trachemys scripta

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. L3:2436 suppresses
  2. L2:1987 excites
  3. L3:1499 weakly suppresses
  4. L3:1510 weakly suppresses
  5. L1:1464 weakly excites
  6. L3:294 weakly excites
  7. L1:1099 weakly excites
  8. L3:2671 weakly suppresses
  9. L3:189 weakly excites
  10. L3:359 weakly suppresses
  11. L1:951 weakly excites
  12. L3:829 weakly suppresses
  13. L3:1224 weakly suppresses
  14. L3:1827 weakly suppresses
  15. L2:3033 weakly excites
  16. L1:1973 weakly suppresses
  17. L1:146 weakly excites
  18. L1:2846 weakly excites
  19. L3:2140 weakly suppresses
  20. L2:483 weakly suppresses
  21. L3:2852 weakly excites
  22. L1:1177 weakly excites
  23. L1:1229 weakly excites
  24. L3:115 weakly suppresses

Feeds later layers

  1. L8:787 excites
  2. L5:1662 excites
  3. L7:1085 excites last digits of big numbers
  4. L6:560 weakly excites
  5. L8:1103 weakly excites
  6. L8:2397 weakly excites
  7. L10:2406 weakly suppresses
  8. L18:1259 weakly suppresses
  9. L15:2476 weakly suppresses
  10. L8:620 weakly suppresses
  11. L13:2568 weakly excites
  12. L24:2545 weakly excites
  13. L8:280 weakly excites
  14. L21:1382 weakly excites
  15. L25:3070 weakly suppresses
  16. L19:2899 weakly excites
  17. L11:1820 weakly suppresses
  18. L6:1099 weakly excites
  19. L9:2006 weakly excites
  20. L16:1606 weakly excites
  21. L23:297 weakly excites
  22. L11:1304 weakly suppresses
  23. L23:2696 weakly suppresses
  24. L11:1215 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.

3 values between -1.19 and -1.002 values between -0.82 and -0.634 values between -0.63 and -0.4482 values between -0.44 and -0.26821 values between -0.26 and -0.072523 values between -0.07 and 0.12449 values between 0.12 and 0.3018 values between 0.30 and 0.491 values between 0.49 and 0.682 values between 0.68 and 0.861 values between 1.61 and 1.791 values between 3.66 and 3.841 values between 6.08 and 6.27offnegativepositiveoften

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. ordo
  2. acre
  3. básica
  4. ollider
  5. rigs
  6. estyle
  7. fetch
  8. 氪
  9. ffects
  10. Sark

Suppresses

  1. 峃
  2. 写作
  3. abs
  4. 绳
  5. integrating
  6. ButtonModule
  7. ichever
  8. integration
  9. :index
  10. writing