Neuron Explorer

Qwen3-0.6B

'and' after action verbs like click or stand

A neuron in layer 9 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 lights up on the word 'and' when it comes right after an action word, as in 'click and begin' or 'stood and proclaimed'. Turning it up made the model's writing wander into vague talk about community and helping others, but the link is loose.

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

    024.⏎The primary purpose behind implementing this law is to ensure that the people of certain industries who got affected because of COVID-19 can come back and work again. This only applies to the employees that were earlier laid off. The

  2. #2 peak

    uan, a Chinese cadre appointed as leader of the Chinese Communist Party in Tibet in March 1992, has called on "inland Chinese to come and help open up Tibet."⏎In 1994 preparatory work began on

  3. #3 peak

    the translation. So if you encounter any problem in our translation service please feel free to correct it at the spot. All you have to do is to click here and submit your correction.⏎Frequently Asked Questions (FAQ)⏎What do you mean by

  4. #4 peak

    handle unlike denominators or how to regroup or to borrow with mixed numbers? Take a deep breath and let us take it one step at a time. Click and begin.⏎GCF and LCM by Prime Factorization⏎A straight forward way

  5. #5 peak

    mentality:⏎– lack of self development;⏎As long as we believe that anything we need would be provided for by someone other than ourselves, we wouldn’t stand up and do something; we would not acquire skills we could use to add value to others

  6. #6 peak

    the failure of countless U.N. resolutions. Their votes were cast after they reviewed the same intelligence data viewed by the president. Their votes were cast after many stood and proclaimed that Iraqi dictator Saddam Hussein must not be allowed to remain in power.⏎

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 2017, the 2018, and the number of topics in the 2018, the number of topics in the "2018". The number of "2018"

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

    In some areas the prevalence of hepatitis C is extremely high, as in Egypt, Saudi Arabia, the Phillipines and Papua New Guinea. The prevalence of hepatitis C antibody in volunteer blood donors is generally estimated at between 0.4% and

  2. #2 peak

    direct clitoral stimulation.⏎Some lesbian and bisexual women object to the term "sc<|endoftext|>Archaeologists have unearthed a hoard of Late Pictish and Roman silver buried in a field in Aberdeenshire.⏎The new finds include

  3. #3 peak

    was born as Ernesto Guevara de la Serna in 1928 to a middle-class family (Castaneda 1998, 3). He was Argentinean by birth but was later awarded with an honorary

  4. #4 peak

    when a load completes a circuit between the two terminals. This load is your device.⏎Why Aren’t All Batteries Rechargeable?⏎Disposable batteries are all alkaline batteries, which means the cathode is made of manganese oxide and the anode

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. L5:2641 excites unclear: scattered word fragments
  2. L6:1679 excites 'and' joining two actions in instructions
  3. L4:57 suppresses
  4. L6:244 suppresses
  5. L6:2747 weakly suppresses
  6. L6:2000 weakly suppresses
  7. L5:1365 weakly suppresses
  8. L2:1669 weakly suppresses
  9. L4:2270 weakly excites
  10. L6:1642 weakly suppresses
  11. L3:262 weakly suppresses
  12. L4:478 weakly suppresses
  13. L6:2407 weakly suppresses
  14. L7:992 weakly excites
  15. L6:1880 weakly excites "by" introducing a deadline or means
  16. L6:2359 weakly excites
  17. L5:1359 weakly suppresses
  18. L6:439 weakly suppresses
  19. L1:1188 weakly excites
  20. L7:138 weakly excites 'to' after verbs like prove or time
  21. L7:57 weakly excites
  22. L3:1177 weakly excites
  23. L1:2825 weakly excites
  24. L5:431 weakly excites

Feeds later layers

  1. L14:1639 excites "and" joining two nouns
  2. L16:376 excites 'or' joining alternatives
  3. L12:2704 excites
  4. L18:2393 weakly excites
  5. L13:1336 weakly suppresses
  6. L14:659 weakly excites
  7. L23:2169 weakly excites
  8. L15:1176 weakly excites
  9. L17:2911 weakly excites
  10. L12:434 weakly excites
  11. L16:1644 weakly suppresses
  12. L16:940 weakly excites
  13. L12:2279 weakly excites
  14. L27:412 weakly suppresses
  15. L10:838 weakly suppresses comma in a list of paired nouns
  16. L24:2377 weakly suppresses
  17. L18:162 weakly excites
  18. L17:989 weakly excites
  19. L19:2040 weakly excites
  20. L17:2521 weakly suppresses
  21. L11:1569 weakly excites
  22. L23:1338 weakly suppresses
  23. L19:617 weakly excites
  24. L13:887 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.

8 values between -0.67 and -0.5643 values between -0.56 and -0.44147 values between -0.44 and -0.33406 values between -0.33 and -0.21879 values between -0.21 and -0.101341 values between -0.10 and 0.01738 values between 0.01 and 0.13204 values between 0.13 and 0.2469 values between 0.24 and 0.3615 values between 0.36 and 0.478 values between 0.47 and 0.5815 values between 0.58 and 0.709 values between 0.70 and 0.812 values between 0.81 and 0.923 values between 0.92 and 1.041 values between 1.04 and 1.151 values between 1.15 and 1.272 values between 1.27 and 1.381 values between 1.38 and 1.491 values between 1.49 and 1.612 values between 1.61 and 1.721 values between 1.72 and 1.842 values between 1.84 and 1.951 values between 1.95 and 2.061 values between 2.06 and 2.182 values between 2.18 and 2.291 values between 2.40 and 2.521 values between 2.86 and 2.971 values between 2.97 and 3.091 values between 3.31 and 3.431 values between 3.66 and 3.771 values between 3.77 and 3.88offnegativepositiveoften

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. Beaut
  2. eva
  3. jom
  4. oulos
  5. redi
  6. olf
  7. ônica
  8. nop
  9. geb
  10. laş

Suppresses

  1. UCE
  2. 在日本
  3. Toe
  4. นโยบาย
  5. пуст
  6. �
  7. inent
  8. 在网络
  9. itian
  10. suspected