Neuron Explorer

Qwen3-0.6B

word-start 'mult' fragment

A neuron in layer 26 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 piece 'mult' at the start of longer words such as 'multivitamin', 'multitudes' or 'multilayer'. Turned down, the model produced a long run of a Chinese counting word, and turned up it wrote loosely on 'the fact that' and similar phrases.

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

    Sickness reliever), Azadirachta indica (Neem), belonging to the Meliaceae family, from prehistoric Indian subcontinent, for its multitudinous medicinal properties (Tanguturi et al. 202

  2. #2 peak

    ''s most intelligent animals. See a great silverback gorilla, the monkey with the biggest nose, how gibbonsswing through the trees, multicolored marmosets, and a day in the life of a gor

  3. #3 peak

    Words with Power, these words slimply leaped up the page:⏎Then again, Christ is constantly associated with the miraculous provision of food. Miracle of feeding large multitudes with very small amounts of food are recorded in all four Gospels, sometimes

  4. #4 peak

    C, 150mg magnesium, 1500mg calcium and 500 mg niacin from dietary sources each day. A good multivitamin/mineral supplement (like Centrum) is also recommended.⏎- Omega

  5. #5 peak

    the effects of high doses of particular vitamins or minerals, like calcium and vitamin D, this clinical trial examined whether a common daily multivitamin had an effect on overall cancer risk. A randomized, double-blinded

  6. #6 peak

    affordable emissions free car The question an issue of engineering and economics is being studied by ... - Read More⏎Researchers at Rice University have created flexible patterned sheets of multilayer graphene from a cheap polymer by burning it with a computer controlled laser The

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:个个个个个个个个个个个个个个个个个个个个个个个个个个个个个个个个个个个个个个个个个个个个个个个个

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

    81 launched a scheme called Fifth Generation, 10-year program for the development of supercomputers was presented in it.⏎Britain created a project called ‘Elvi’ whereas European Union countries also introduced a program called ‘Esprit’.⏎

  2. #2 peak

    If exp is 0, x is returned.⏎These functions will fail if:⏎The result overflows.⏎If the integer expression (math_errhandling & MATH_ERREXCEPT) is non-zero, the overflow floating-point exception is raised.⏎

  3. #3 peak

    of the commoner.⏎This phrase originates from Rome in Satire X of the Roman satirist and poet Juvenal (circa A.D. 100). In context, the Latin metaphor panem et circenses (bread

  4. #4 peak

    activate subtitles in english for better understanding.⏎John Forbes Nash Jr., a mathematical genius whose struggle with schizophrenia was chronicled in the 2001 movie “A

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. L23:2883 strongly excites
  2. L22:2314 excites
  3. L9:59 suppresses "a" meaning per (a day, a year)
  4. L4:326 suppresses
  5. L11:233 suppresses
  6. L19:1935 excites
  7. L7:224 suppresses
  8. L20:311 suppresses
  9. L17:1818 excites
  10. L21:708 suppresses
  11. L4:961 excites
  12. L4:205 excites the digit 2 inside citations
  13. L16:600 suppresses
  14. L18:973 excites
  15. L21:1863 suppresses
  16. L14:2220 weakly excites "this" pointing back to a thing
  17. L4:731 weakly excites
  18. L18:2629 weakly suppresses
  19. L18:2804 weakly suppresses
  20. L24:2344 weakly suppresses
  21. L9:13 weakly excites
  22. L19:3060 weakly suppresses
  23. L18:2183 weakly suppresses
  24. L18:635 weakly excites

Feeds later layers

  1. L27:1694 weakly excites
  2. L27:622 weakly excites
  3. L27:994 weakly suppresses
  4. L27:653 weakly suppresses
  5. L27:2365 weakly suppresses
  6. L27:1908 weakly excites the 'd' that starts 'die' or 'dead'
  7. L27:2701 weakly excites
  8. L27:925 weakly excites
  9. L27:2744 weakly suppresses
  10. L27:2743 weakly suppresses
  11. L27:1522 weakly excites
  12. L27:1467 weakly suppresses
  13. L27:3017 weakly suppresses
  14. L27:2271 weakly excites
  15. L27:1782 weakly suppresses
  16. L27:1276 weakly suppresses
  17. L27:1425 weakly suppresses
  18. L27:2599 weakly suppresses
  19. L27:624 weakly suppresses
  20. L27:193 weakly excites
  21. L27:2393 weakly suppresses
  22. L27:703 weakly suppresses
  23. L27:1544 weakly excites
  24. L27:469 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.

1 values between -20.27 and -19.251 values between -14.18 and -13.171 values between -13.17 and -12.152 values between -11.14 and -10.133 values between -9.11 and -8.101 values between -8.10 and -7.081 values between -7.08 and -6.076 values between -6.07 and -5.069 values between -5.06 and -4.0413 values between -4.04 and -3.0324 values between -3.03 and -2.0134 values between -2.01 and -1.001078 values between -1.00 and 0.022650 values between 0.02 and 1.0344 values between 1.03 and 2.0415 values between 2.04 and 3.066 values between 3.06 and 4.072 values between 4.07 and 5.095 values between 5.09 and 6.102 values between 6.10 and 7.111 values between 7.11 and 8.132 values between 8.13 and 9.141 values between 11.17 and 12.181 values between 12.18 and 13.202 values between 15.23 and 16.242 values between 17.26 and 18.271 values between 19.28 and 20.30offnegativepositiveoften

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. illance
  2. ellation
  3. ishments
  4. acle
  5. ulary
  6. ivals
  7. /cms
  8. utt
  9. ishing
  10. AMERA

Suppresses

  1. 个
  2. 个小
  3. 种
  4. 些
  5. 個
  6. 个工作
  7. 个城市
  8. 个多
  9. 个国家
  10. 个项目