Neuron Explorer

Qwen3-0.6B

second halves of split names and acronyms

A neuron in layer 21 of Qwen3-0.6B. Its strongest snippets are mixed: they peak on different tokens. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on the later pieces of a name or abbreviation that the tokenizer has split into parts, such as the 'lu' in RLUIPA or the 'M' in MPA. Turning it up led to unrelated text, so the steering results do not add much.

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

    houses of worship, and religious schools. The full text of RLUIPA is available at http://www.usdoj.gov/crt/housing/housing_rluipa.htm. Below is a summary of the law’s key provisions, with illustrations

  2. #2 peak

    religion?⏎Íocaíocht & Slándáil⏎Déantar do chuid sonraí íocaíochta a phróiseáil go slán. Ní dhéanann muid sonraí cártaí cre

  3. #3 peak

    known as Kurd Dagh (Çiyayê Kurd in Kurdish), meaning Mountain of the Kurds. The major urban center is Efrîn (‘Afrin in Arabic), with an urban population of approximately 80,00

  4. #4 peak

    distribution and host preference of lianas in a tropical dry evergreen forest in India.⏎Vivek, P., & Parthasarathy, N.⏎Lianas are structurally dependent life-form that rely mostly on trees to access the forest

  5. #5 peak

    recorded with various hanja phonetically approximating its native Korean name: 斯盧 (사로, Saro), 斯羅 (사라, Sara), 徐那(伐) (서나[벌], Seona

  6. #6 peak

    and skill level different and above that for⏎passenger cars, particularly when the vans are fully loaded or drivers⏎experience an emergency situation.”3 While the NTSB had⏎previously made recommendations about driver training and use policies⏎for fifteen

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 way to the present in the way of the real, the way of the real that is the by the way of the real, the by the way of the real, by the way of the real, the by the way of 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

    electronic forms of online payment.⏎9. Be on guard against scams hiding behind online coupon offers.⏎10. Be careful if you download banking software onto

  2. #2 peak

    but their design limitations make them of little value for risk evaluation⏎- Assessing whether fluoride constitutes a risk factor for osteosarcoma is complicated by the rarity of the disease and the difficulty of characterizing biologic dose because of the ubiquity

  3. #3 peak

    ism was shown off at the 1933 Radio Exhibition at the Olympia exhibition space in London. The most intriguing aspect of its design is the world map sitting just below the screen. One

  4. #4 peak

    effort to prove we could create a commercial quality game using NASA content that is fun. NASA was looking for a project at the same time the Army Game Studio was looking to branch out from America’s Army. It helped that Army Game Studio and NASA

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. L18:569 suppresses
  2. L0:1224 excites
  3. L6:13 suppresses
  4. L0:1301 excites
  5. L8:7 suppresses
  6. L0:178 excites
  7. L8:10 suppresses
  8. L2:75 excites
  9. L7:110 excites
  10. L1:1464 suppresses
  11. L6:7 excites
  12. L15:82 excites
  13. L13:983 weakly suppresses
  14. L5:46 weakly suppresses
  15. L16:2176 weakly excites
  16. L11:322 weakly excites
  17. L5:50 weakly suppresses
  18. L5:169 weakly suppresses
  19. L3:56 weakly suppresses
  20. L4:97 weakly excites
  21. L11:456 weakly excites
  22. L5:110 weakly excites
  23. L16:1392 weakly excites
  24. L8:4 weakly suppresses

Feeds later layers

  1. L27:188 excites
  2. L27:206 excites
  3. L26:214 excites
  4. L22:80 excites
  5. L22:85 excites
  6. L22:84 excites
  7. L26:242 excites
  8. L22:99 excites
  9. L22:79 excites
  10. L22:66 excites
  11. L22:982 suppresses
  12. L22:128 excites
  13. L22:112 excites
  14. L23:44 excites
  15. L24:1017 suppresses
  16. L22:92 excites
  17. L22:58 weakly excites
  18. L22:269 weakly excites
  19. L22:88 weakly excites
  20. L22:90 weakly excites
  21. L22:98 weakly excites
  22. L22:59 weakly excites
  23. L23:666 weakly suppresses
  24. L22:337 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.

2 values between -14.64 and -13.642 values between -13.64 and -12.641 values between -12.64 and -11.641 values between -11.64 and -10.653 values between -10.65 and -9.6510 values between -9.65 and -8.6513 values between -8.65 and -7.6517 values between -7.65 and -6.6527 values between -6.65 and -5.6547 values between -5.65 and -4.6588 values between -4.65 and -3.65110 values between -3.65 and -2.66159 values between -2.66 and -1.66204 values between -1.66 and -0.66304 values between -0.66 and 0.34323 values between 0.34 and 1.34370 values between 1.34 and 2.34376 values between 2.34 and 3.34386 values between 3.34 and 4.34376 values between 4.34 and 5.34308 values between 5.34 and 6.34240 values between 6.34 and 7.33171 values between 7.33 and 8.33125 values between 8.33 and 9.3383 values between 9.33 and 10.3365 values between 10.33 and 11.3334 values between 11.33 and 12.3327 values between 12.33 and 13.3317 values between 13.33 and 14.338 values between 14.33 and 15.325 values between 15.32 and 16.321 values between 18.32 and 19.322 values between 19.32 and 20.321 values between 20.32 and 21.321 values between 22.32 and 23.321 values between 24.31 and 25.31offnegativepositiveoften

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. UserController
  3. ToProps
  4. encodeURIComponent
  5. 忘れ
  6. eventType
  7. URLConnection
  8. strawberries
  9. Controllers
  10. addObject

Suppresses

  1. ucle
  2. andal
  3. ktor
  4. IFE
  5. ine
  6. лиз
  7. orado
  8. bia
  9. eload
  10. лен