Neuron Explorer

Qwen3-0.6B

unclear: mid-word pieces like 'sup' and 'mad'

A neuron in layer 25 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 small chunks of longer words, such as the 'sup' in 'supplied' or 'supersonic' and the 'Mad' in 'Madinah'. It does not seem to care about the topic. Turning it up made the model produce loose, repetitive text with no clear theme.

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

    foremost combat weapons.⏎Acknowledging the significance of the availability of good shields for successful warfare, Carolingian rulers took special care to ensure that their soldiers were well-supplied with ample shields.⏎Shields were used by both the cavalrymen and infantry

  2. #2 peak

    North and Central America to provide population data for use in science and conservation biology and is the longest-running citizen survey in the world.⏎The Matagorda County Mad Island Marsh Christmas Bird Count (CBC) is one of the most successful in the

  3. #3 peak

    pools or water parks.⏎- Use of Tampons⏎- Problems emptying the bladder completely⏎- Having a urinary catheter⏎- Kidney stones⏎- Suppressed immune system⏎- Immobility for a long period⏎- Holding your

  4. #4 peak

    rows;⏎indeed, such sorrows were his bosom companions⏎in that frozen island dungeon⏎where Nithad fettered him:⏎so many strong-but-supple sinew-bands⏎binding the better man.⏎That passed away; this

  5. #5 peak

    the 14,000-square kilometer Harrat Khaybar, located approximately 137 kilometers to the northeast of the city of Al Madinah (Medina). According to scientists, the volcanic field was formed by

  6. #6 peak

    be the biggest of all dog breeds around the globe. They’re renowned for their majesty and imposing presence. While they are often gentle in nature and even mild-mannered members of many families, they’re still animals. Just like any dog

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 use the photos photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo photo

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

    and not appearing to be surface feeding at all.’⏎- ‘And wigeon and scaup, teal and shovelers, buffleheads and a few mottled ducks.’⏎- ‘The sea ducks - common golden-eyes and

  2. #2 peak

    his neighbours demonstrated that diversification of seeds and tubers, along with traditional methods of preparing the soil, enhanced crop and grassland yields. Although the region is economically impoverished, he showed that by reviving the diversity of their natural heritage, rather

  3. #3 peak

    used mortality, and isolation of small populations are major concerns in the conservation of large carnivores (Clark et al., 1996; Singleton & Lehmkuhl, 2001). The loss of habitat and connectivity between

  4. #4 peak

    stood out from the majority of artists living there. First, she was a single young female in a society that basically frowned upon such status. Second, she was financially independent, and third, she had received a typical "proper" education as beh

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. L24:1190 excites
  2. L21:2519 suppresses
  3. L22:1138 suppresses
  4. L23:1046 suppresses
  5. L21:1679 excites
  6. L16:1408 weakly suppresses
  7. L19:1373 weakly suppresses
  8. L23:206 weakly suppresses
  9. L19:2007 weakly excites
  10. L2:630 weakly suppresses
  11. L18:2383 weakly suppresses
  12. L12:1214 weakly excites
  13. L5:2146 weakly suppresses
  14. L21:1047 weakly excites
  15. L3:2045 weakly suppresses
  16. L4:1636 weakly excites
  17. L20:1319 weakly suppresses
  18. L9:1713 weakly suppresses
  19. L14:693 weakly excites
  20. L8:2055 weakly excites
  21. L22:151 weakly suppresses
  22. L0:55 weakly excites
  23. L18:328 weakly suppresses
  24. L14:2759 weakly suppresses

Feeds later layers

  1. L27:873 weakly excites
  2. L26:2600 weakly excites
  3. L26:2476 weakly suppresses
  4. L27:1899 weakly suppresses
  5. L26:2792 weakly suppresses
  6. L27:2244 weakly suppresses
  7. L27:2031 weakly excites
  8. L27:996 weakly excites
  9. L27:168 weakly excites
  10. L27:544 weakly excites
  11. L27:1878 weakly suppresses
  12. L27:1830 weakly excites
  13. L26:558 weakly suppresses
  14. L27:2796 weakly suppresses
  15. L26:1897 weakly suppresses
  16. L26:1938 weakly excites
  17. L26:2319 weakly excites
  18. L26:2126 weakly excites
  19. L27:2546 weakly excites
  20. L26:2321 weakly excites
  21. L26:2875 weakly excites
  22. L27:1206 weakly suppresses
  23. L26:1626 weakly excites the word 'I' as subject
  24. L26:1812 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 -37.59 and -35.641 values between -31.72 and -29.761 values between -25.85 and -23.891 values between -23.89 and -21.931 values between -21.93 and -19.971 values between -19.97 and -18.022 values between -16.06 and -14.103 values between -14.10 and -12.145 values between -12.14 and -10.1812 values between -10.18 and -8.2315 values between -8.23 and -6.2726 values between -6.27 and -4.3147 values between -4.31 and -2.35515 values between -2.35 and -0.403181 values between -0.40 and 1.5646 values between 1.56 and 3.5216 values between 3.52 and 5.489 values between 5.48 and 7.447 values between 7.44 and 9.395 values between 9.39 and 11.353 values between 11.35 and 13.311 values between 13.31 and 15.272 values between 15.27 and 17.232 values between 17.23 and 19.181 values between 19.18 and 21.141 values between 27.01 and 28.971 values between 28.97 and 30.931 values between 30.93 and 32.891 values between 38.76 and 40.72offnegativepositiveoften

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. <Option
  2. 日消息
  3. offs
  4. undert
  5. weekends
  6. eft
  7. paid
  8. yme
  9. ERSION
  10. _axes

Suppresses

  1. minster
  2. documentation
  3. mine
  4. edly
  5. ⎣
  6. converged
  7. dataType
  8. omb
  9. monton
  10. fony