Neuron Explorer

Qwen3-0.6B

"enough" before "to"

A neuron in layer 20 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 word "enough" in phrases like "strong enough to withstand" or "not enough heat to", where the next word is usually "to". Turning it down made the model repeat "right enough enough", which shows the link to this word, while turning it up gave unrelated text.

Written by Claude from this neuron's activations and steering outputs. Confidence: high. 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

    metal mix together. From there, it goes in the sintering furnace.⏎The sintering furnace applies enough heat to further bond the particles -- but not enough heat to reach the melting point of your alloying elements.⏎The steps in powder

  2. #2 peak

    smoking, fast food and the remote control.⏎Among 22 mummies who received full-body computed tomography scans, 16 had hearts or arteries preserved enough to study. Of those, nine had evidence of blockage from atherosclerosis

  3. #3 peak

    stay calm. We need to listen to what they are saying—really listen. And as we hear their arguments we learn to set aside our own defensiveness enough to note

  4. #4 peak

    help from the furniture from just before the end of their first year, and finally walk with some help. They will bring objects to you, having built their confidence enough to carry things and interact.⏎-

  5. #5 peak

    may have been affected by other contextual factors. The economic rationale for FFE programs is to offer free food conditional on school attendance to increase the net benefits of schooling enough to change some households' decisions about

  6. #6 peak

    aluminum framework. Figure 5 shows a typical enclosure, which is elevated above ground to prevent flooding. The enclosure is designed to be corrosion and UV resistant and strong enough to withstand falling trees. The electronics within the enclosure are shown in Figure 6

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 right enough enough enough enough enough right enough enough enough right enough enough enough right enough enough enough right enough enough enough right enough enough enough right enough enough enough right enough enough enough right enough enough enough right enough enough enough right enough enough enough right

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

    your professional development<|endoftext|>Mackinder’s Heartland Theory made SIMPLE⏎Do you want to learn about Mackinder’s Heartland Theory? Then you have come to the right

  2. #2 peak

    to be the co-principal<|endoftext|>Choosing the Right Electrode for pH Measurements⏎Tip: Correct and accurate pH measurements depend upon reliable instruments and electrodes. Selecting the right equipment and procedures for handling and maintenance are crucial for obtaining optimum results and

  3. #3 peak

    system and ensures it operates efficiently, saving energy and reducing utility costs.⏎Choosing the Right Replacement Parts⏎When it comes to repairing or upgrading an HVAC system, choosing the correct replacement parts is critical. Quality parts ensure compatibility with the system and extend its

  4. #4 peak

    of problems in those that suffer from stomach sensitivities.⏎If you’re interested in learning a little more about buckwheat, rest assured that you have come to the right place.⏎In this

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. L17:2863 strongly excites
  2. L18:2733 suppresses
  3. L19:528 suppresses
  4. L19:797 suppresses
  5. L19:2954 suppresses
  6. L15:1105 excites the word 'enough' and similar
  7. L15:2661 suppresses
  8. L19:3025 excites
  9. L18:3055 excites
  10. L19:945 suppresses
  11. L12:1414 suppresses
  12. L13:2838 suppresses
  13. L12:2818 excites
  14. L15:2315 excites
  15. L12:244 suppresses
  16. L16:1806 suppresses
  17. L11:1554 weakly suppresses
  18. L13:1949 weakly suppresses
  19. L14:1481 weakly excites
  20. L17:2263 weakly excites
  21. L17:157 weakly suppresses
  22. L12:468 weakly suppresses
  23. L10:2526 weakly excites
  24. L16:2961 weakly excites

Feeds later layers

  1. L22:1234 strongly suppresses
  2. L23:2442 strongly suppresses
  3. L24:1858 strongly suppresses
  4. L27:1076 suppresses
  5. L24:1912 suppresses
  6. L27:1489 suppresses
  7. L23:1829 suppresses
  8. L23:1038 suppresses
  9. L21:1199 suppresses
  10. L22:1028 suppresses
  11. L24:2196 suppresses
  12. L23:2383 suppresses
  13. L23:1113 suppresses
  14. L23:1418 weakly suppresses
  15. L21:870 weakly suppresses
  16. L21:2956 weakly suppresses
  17. L22:1817 weakly suppresses
  18. L21:1482 weakly suppresses
  19. L23:971 weakly suppresses
  20. L24:1377 weakly suppresses
  21. L27:778 weakly suppresses
  22. L24:64 weakly suppresses
  23. L21:2363 weakly suppresses
  24. L21:2415 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 -13.10 and -12.641 values between -9.38 and -8.921 values between -8.92 and -8.451 values between -8.45 and -7.991 values between -7.06 and -6.592 values between -6.13 and -5.661 values between -5.19 and -4.733 values between -4.73 and -4.262 values between -4.26 and -3.803 values between -3.80 and -3.333 values between -3.33 and -2.871 values between -2.87 and -2.405 values between -2.40 and -1.949 values between -1.94 and -1.4713 values between -1.47 and -1.0142 values between -1.01 and -0.54491 values between -0.54 and -0.082378 values between -0.08 and 0.39816 values between 0.39 and 0.85116 values between 0.85 and 1.3210 values between 1.32 and 1.784 values between 1.78 and 2.252 values between 2.71 and 3.181 values between 3.18 and 3.641 values between 5.04 and 5.50offnegativepositiveoften

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. lesc
  2. arde
  3. ician
  4. ída
  5. lesi
  6. 鹂
  7. ardo
  8. orney
  9. 嗓子
  10. 来讲

Suppresses

  1. 正确的
  2. enough
  3. 正确
  4. correct
  5. correct
  6. sufficient
  7. Correct
  8. proper
  9. 足够
  10. correctly