Neuron Explorer

Qwen3-0.6B

'the' before river names

A neuron in layer 23 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 word 'the' when a river, or a body of water, is about to be named, such as 'join the Seine' or 'banks of the Mississippi'. It stays quiet on most other uses of 'the'. Turning it up gave odd text with no clear link to rivers.

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

    enters France northeast of Hirson, 9 miles (15 km) from its source, and flows generally southwestward, watering the Paris Basin, to join the Seine River at Conflans, after a course of 188

  2. #2 peak

    ). It is also one of the “finest detours in France” network (Plus beaux détours de France). Stretching out along the banks of the River Ornain, it was, for many years, the capital of the

  3. #3 peak

    field, was the birthplace of modern Manchester.⏎It was established by 78 AD at a site overlooking the confluence of the rivers Irwell and Medlock.⏎This was a timber fort with earthen ramp

  4. #4 peak

    of Saint-Gobain, Compiègne, Hallatte, and Chantilly. Near Compiègne it receives its main tributary, the Aisne; and below the industrial town of Creil, it is joined

  5. #5 peak

    en gravure kaart] Schenckenschans, 1635.⏎Plan of the fortification of Schenkenschans, located between the rivers Rhine and Waal, including details on the landscape. During the Eighty

  6. #6 peak

    3, and 4 (teen team) will work between the remaining walls of a Roman villa on Poggio del Molino (Windmill Hill), overlooking the Mediterranean Sea. The villa was built at the start of the 1st century

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: how to test a real-time API that is only public in a private network only when you need to, and only when you know that you are only testing, and that you know that you are only testing. I thought that this might be

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

    Magnet or Ferromagnetic Material. The Stepper Motors are manufactured with steps per revolution of 12, 24, 72, 144, 180, and 200, resulting in stepping

  2. #2 peak

    that they were propelled primarily or entirely by sail. The length of the round ships ranged from 26 to 48 feet (8 to 14.7 meters), and they were between 8 and 16 feet (2

  3. #3 peak

    of them 10,000 km (6200 miles) away, and one 18,000 km (11,100 miles) from the blast. These stations monitor seismic waves, infr

  4. #4 peak

    said Lopez at the site, where the temperature reaches 45 C (113 F) at this time of year.⏎The first investigatory work was done

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. L5:1546 excites
  2. L20:1308 suppresses the digit 3 inside years
  3. L20:361 excites
  4. L20:931 weakly suppresses the digit 6 inside years
  5. L1:1556 weakly excites
  6. L11:223 weakly suppresses
  7. L17:30 weakly suppresses
  8. L17:106 weakly excites
  9. L17:213 weakly excites
  10. L18:750 weakly suppresses
  11. L21:614 weakly excites
  12. L4:498 weakly excites
  13. L5:676 weakly excites
  14. L19:427 weakly suppresses
  15. L10:1349 weakly excites
  16. L16:177 weakly suppresses zero digit inside dates and codes
  17. L15:1452 weakly suppresses
  18. L2:2647 weakly suppresses
  19. L14:1541 weakly excites
  20. L18:242 weakly suppresses
  21. L16:75 weakly excites
  22. L19:1269 weakly suppresses the word 'about'
  23. L12:2888 weakly excites
  24. L20:3023 weakly excites

Feeds later layers

  1. L24:1968 strongly excites
  2. L25:2442 excites
  3. L24:2840 suppresses
  4. L26:1485 weakly excites
  5. L26:1221 weakly suppresses
  6. L24:1460 weakly suppresses
  7. L26:697 weakly excites
  8. L27:2417 weakly excites
  9. L27:1473 weakly suppresses
  10. L25:1974 weakly excites
  11. L24:2466 weakly suppresses
  12. L27:2151 weakly suppresses
  13. L26:1273 weakly suppresses
  14. L27:586 weakly excites
  15. L26:200 weakly excites
  16. L26:2915 weakly excites
  17. L25:1796 weakly excites
  18. L27:127 weakly suppresses
  19. L25:2345 weakly excites
  20. L27:2444 weakly suppresses
  21. L27:2852 weakly excites
  22. L26:1621 weakly suppresses
  23. L26:1253 weakly excites
  24. L25:1598 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 -18.94 and -17.891 values between -11.61 and -10.561 values between -9.51 and -8.472 values between -7.42 and -6.373 values between -6.37 and -5.3210 values between -5.32 and -4.289 values between -4.28 and -3.2312 values between -3.23 and -2.1844 values between -2.18 and -1.131423 values between -1.13 and -0.092251 values between -0.09 and 0.9673 values between 0.96 and 2.0120 values between 2.01 and 3.0624 values between 3.06 and 4.1010 values between 4.10 and 5.157 values between 5.15 and 6.205 values between 6.20 and 7.242 values between 7.24 and 8.292 values between 8.29 and 9.341 values between 9.34 and 10.395 values between 11.43 and 12.481 values between 13.53 and 14.581 values between 21.91 and 22.95offnegativepositiveoften

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. odies
  2. 上的
  3. odem
  4. abel
  5. 呼和
  6. asses
  7. 上の
  8. asher
  9. 意义上
  10. thes

Suppresses

  1. leccion
  2. warts
  3. SPATH
  4. endar
  5. thought
  6. eguard
  7. prend
  8. ppelin
  9. umpy
  10. orias