Neuron Explorer

Qwen3-0.6B

the gap in number ranges

A neuron in layer 15 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 fires on the spot between two numbers in a range, like the gap in '10-15', '50 to 100' or '12 to 18 inches'. It is most active right before the second number of the range is about to appear. When turned up the model produced repeated words or long strings of zeros, so the steering output does not show what the neuron does.

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

    quite variable, but a reasonable mean is about 10-15 from each protein. They all have to agree in their "chemical properties

  2. #2 peak

    this, he proposed a machine that can collect data, compiling it into a type of artificial intelligence.⏎"I think as we look toward the next 50 to 100 years, if we really can come up with these machines that can

  3. #3 peak

    't known how long they can live under professional care, she said, because zoos and aquariums have only tracked animal ages for the last 30 to 40 years.⏎Thunder, who was most recently seen in the Canyon Falls exhibit

  4. #4 peak

    risk for type 2 diabetes, heart disease, osteoarthritis, and certain forms of cancer.⏎âIf youâre overweight, losing just 5 to 10 percent of your body weight can slash the risk of both type

  5. #5 peak

    torso in horizontal line, rotating shoulders. Head and shoulders stay aligned so they move as single unit. Partners reverse sequence and return to start, performing 10–15 reps before switching sides.⏎Medicine Ball Partner Twists⏎Partners

  6. #6 peak

    full sun and well-drained soil. Prepare the garden bed by using a garden fork or tiller to loosen the soil to a depth of 12 to 15 inches, then mix in a 2- to 4-inch layer

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 have, in my life, as, and, and, and, and, and, and, and, and, as, and, and, and, and, and, and, and, and, and, and, and

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

    €™s WONDER database, which tracks mortality data and causes of death in the U.S. Researchers first isolated all opioid-related deaths recorded between 2001 and 2016 (335,1

  2. #2 peak

    Large Scale Biosphere-Atmosphere|⏎|Time Period||1985-01-01 to 2001-12-

  3. #3 peak

    6 and ES-3, tbl. 7, with EPA, Inventory of U.S. Greenhouse Gas Emissions and Sinks: 1990-2019<|endoftext|>Following directions with color words is a fun practice activity

  4. #4 peak

    .⏎Clinics that gave the suspect shots<|endoftext|>Classic DACB CollectionAll articles created or submitted in the first twenty years of the project, from 1995 to 2015.⏎Jacob Osasona was born 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. L13:1435 excites
  2. L11:2354 suppresses
  3. L12:947 suppresses
  4. L5:1401 weakly excites
  5. L13:2019 weakly excites
  6. L7:1643 weakly suppresses
  7. L12:1965 weakly suppresses
  8. L11:2811 weakly suppresses
  9. L1:2559 weakly excites
  10. L13:2379 weakly suppresses the digit zero in numbers
  11. L13:707 weakly suppresses
  12. L8:932 weakly suppresses
  13. L9:850 weakly suppresses
  14. L5:778 weakly suppresses
  15. L10:2889 weakly excites
  16. L8:2001 weakly excites
  17. L8:1946 weakly suppresses comma after 'who' or 'which'
  18. L11:1878 weakly excites
  19. L3:2724 weakly suppresses
  20. L8:564 weakly suppresses
  21. L8:364 weakly suppresses
  22. L7:1429 weakly excites
  23. L6:1577 weakly excites
  24. L14:1420 weakly suppresses

Feeds later layers

  1. L16:808 strongly excites
  2. L18:1413 excites
  3. L18:2454 excites
  4. L17:483 suppresses
  5. L18:2238 excites
  6. L18:1207 excites
  7. L21:1972 excites
  8. L17:2704 excites
  9. L19:1393 excites
  10. L16:548 excites
  11. L21:1857 excites
  12. L18:2362 excites
  13. L19:2835 excites
  14. L19:659 excites
  15. L23:223 excites
  16. L22:1248 excites
  17. L17:1468 excites
  18. L19:2735 excites
  19. L25:41 suppresses
  20. L27:376 excites
  21. L20:1009 excites
  22. L17:2116 excites
  23. L20:560 excites
  24. L20:742 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.

3 values between -3.86 and -3.521 values between -3.52 and -3.191 values between -3.19 and -2.851 values between -2.85 and -2.511 values between -2.51 and -2.183 values between -2.18 and -1.845 values between -1.84 and -1.518 values between -1.51 and -1.1712 values between -1.17 and -0.8451 values between -0.84 and -0.50786 values between -0.50 and -0.172540 values between -0.17 and 0.17409 values between 0.17 and 0.5126 values between 0.51 and 0.8416 values between 0.84 and 1.1812 values between 1.18 and 1.5110 values between 1.51 and 1.856 values between 1.85 and 2.182 values between 2.18 and 2.521 values between 2.52 and 2.852 values between 2.85 and 3.192 values between 3.19 and 3.522 values between 3.52 and 3.861 values between 3.86 and 4.202 values between 5.20 and 5.542 values between 5.54 and 5.871 values between 7.89 and 8.221 values between 8.56 and 8.891 values between 9.23 and 9.56offnegativepositiveoften

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. 或
  2. or
  3. 或者是
  4. または
  5. 或多
  6. 或
  7. 或是
  8. hoặc
  9. /she
  10. 或者

Suppresses

  1. chair
  2. rek
  3. ums
  4. ceu
  5. rix
  6. uner
  7. enade
  8. 法
  9. 법
  10. AMPLE