Neuron Explorer

Qwen3-0.6B

"to" after "it's important"

A neuron in layer 1 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 "to" when it comes right after phrases like "it's important" or "remember", where a tip or advice is about to follow. When it was turned up, the model's replies did not show a clear theme.

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

    anged wheel⏎When evaluating wheel products for a specific application, it's important to take into consideration all possible factors that will affect performance. We discuss a few potential

  2. #2 peak

    tight, you’re wrong.”⏎While this isn’t wrong advice, it’s important to understand where this tourniquet procedure came from and why we’re constantly being

  3. #3 peak

    out any forms, or give out money or credit card details.⏎- Don't accept any offers that seem too good to be true - they probably are.⏎Remember to always be

  4. #4 peak

    wheel. It's also important to include a safety variable for any unforeseen circumstances. We recommend a minimum variable of

  5. #5 peak

    compacted or poorly drained. Instead of treating the symptoms using pesticides, it’s important to discover and treat the reason the<|endoftext|>As the summer comes, families need to cool

  6. #6 peak

    nationwide. But it’s also important to remember one of the most vital purposes of offering school meals: fighting hunger so kids

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: I have been going to the moon, and I have been going to the moon again. I have been going to the moon twice, and I have been going to the moon three. I have been going to the moon four. I have

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

    calories, which helps you lose weight. If calories consumed in food are less than calories spent during exercise and other daily activities, you will lose weight. It is impossible not to! Many people have learned how to lose belly fat by starting a running

  2. #2 peak

    more than one edge. So when I draw this graph on my 2D screen, it can happen that edges will be crossing each other.⏎It is impossible to avoid crossing edges, but I guess there is some algorithm which is able to

  3. #3 peak

    Kennedy Shelley⏎Governors in several states have closed down beaches in order to promote social distancing during the 2020 Coronavirus scare.⏎It is impossible to find a tanning bed place that is still open while everyone is COVID-

  4. #4 peak

    familiar with the condition and don’t know what to look for. You may want to consider getting a second opinion if your symptoms do not respond.<|endoftext|>It is impossible to think of the Guatemalan, Mexican and Mesoamerican diet without maize

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. L0:1225 suppresses
  2. L0:1513 weakly suppresses
  3. L0:2216 weakly excites
  4. L0:3024 weakly suppresses
  5. L0:884 weakly excites
  6. L0:2237 weakly excites
  7. L0:2006 weakly suppresses
  8. L0:2923 weakly suppresses
  9. L0:1301 weakly suppresses
  10. L0:1832 weakly excites
  11. L0:1202 weakly suppresses
  12. L0:2812 weakly suppresses
  13. L0:1763 weakly excites
  14. L0:178 weakly suppresses
  15. L0:2408 weakly excites
  16. L0:1239 weakly excites
  17. L0:1889 weakly suppresses
  18. L0:2808 weakly suppresses
  19. L0:977 weakly suppresses "like" after "you'd" or "something"
  20. L0:1907 weakly excites
  21. L0:886 weakly suppresses
  22. L0:1918 weakly suppresses
  23. L0:1198 weakly excites
  24. L0:1985 weakly suppresses

Feeds later layers

  1. L7:1056 weakly excites
  2. L13:959 weakly suppresses
  3. L10:1753 weakly suppresses
  4. L11:1860 weakly excites
  5. L3:509 weakly excites
  6. L14:1878 weakly suppresses
  7. L13:1711 weakly suppresses
  8. L4:1239 weakly suppresses
  9. L20:942 weakly excites
  10. L2:2747 weakly suppresses
  11. L25:2307 weakly suppresses
  12. L16:2397 weakly suppresses
  13. L16:744 weakly excites
  14. L15:2385 weakly suppresses
  15. L5:2893 weakly suppresses
  16. L12:452 weakly excites
  17. L17:1822 weakly excites 'should' in advice and claims
  18. L21:2823 weakly suppresses
  19. L2:494 weakly suppresses
  20. L13:2663 weakly excites
  21. L5:1883 weakly excites
  22. L8:1465 weakly excites
  23. L3:1030 weakly excites
  24. L9:3021 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 -1.98 and -1.802 values between -0.92 and -0.755 values between -0.75 and -0.5712 values between -0.57 and -0.4065 values between -0.40 and -0.22880 values between -0.22 and -0.052535 values between -0.05 and 0.13346 values between 0.13 and 0.3136 values between 0.31 and 0.489 values between 0.48 and 0.664 values between 0.66 and 0.833 values between 0.83 and 1.014 values between 1.01 and 1.183 values between 1.18 and 1.361 values between 2.59 and 2.761 values between 3.82 and 3.991 values between 4.87 and 5.05offnegativepositiveoften

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. eyed
  2. quên
  3. 烊
  4. усп
  5. geom
  6. geometric
  7. äm
  8. canf
  9. precaution
  10. stand

Suppresses

  1. """,⏎
  2. _MEDIUM
  3. ilians
  4. rack
  5. geries
  6. iationException
  7. hält
  8. _trampoline
  9. alie
  10. illo