Neuron Explorer

Qwen3-0.6B

the word 'without'

A neuron in layer 24 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 "without" when it comes after a clause, as in "without fail", "without doubt" or "without warning". Turning it down or up did not change the generated text in any obvious way.

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

    intent should never be considered, ever. To even look down that road is foolish.⏎But consider this: if you walk into any contemporary art museum you will, without fail, find on the wall next to the art, a statement from the artist

  2. #2 peak

    .<|endoftext|>Though buckwheat is certainly one of the lesser known healthy grains out there, we’re sure that you’ll have had it at least once or twice without even

  3. #3 peak

    dance has had the ability to transcend all borders of language, geography and age. It is a primal urge, a surge of energy, an uplifting feeling and, without doubt, a gift.<|endoftext|>June 8, 2000⏎

  4. #4 peak

    <|endoftext|>The eyes are the windows to the soul, or so goes the proverb. As humans, we use our eyes to express a wide variety of emotions, often without even realizing we’re doing

  5. #5 peak

    this method will help you achieve a bouncing ball animation in probably 5 minutes or less.⏎So without wasting much time, let’s get into it!⏎Here are the steps:⏎-

  6. #6 peak

    at the head of the bay behind several small islands. It was a nice sheltered spot. It would even be safe from the occasional sudden great storms that struck without warning in that late summer season. As the men heaved in the anchor and

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: If the fact that certain parts of the world at certain time periods about at certain places at certain things at certain things about certain things at certain time periods about at certain time periods about at certain time periods about at certain time periods about at certain

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

    enteritis. A bacteria called clostridia was isolated as the organism responsible.⏎Necrotizing enteritis is associated with poor hygiene and an inadequate diet, and it was found that

  2. #2 peak

    γ-proteobacterium, W. melaphagi is a rhizobacterium).⏎Being present endosymbiotically in 20+% of

  3. #3 peak

    Library Board Images Online⏎Anthony Farrington Trading Places, The East India Company and Asia 1600-

  4. #4 peak

    |• Land||576 sq mi (1,490 km2)|⏎|• Water||83 sq mi (210 km2) 13%%|⏎|• Density||401/sq mi

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. L22:390 excites
  2. L20:692 suppresses
  3. L20:1414 weakly excites
  4. L8:2442 weakly suppresses
  5. L4:1711 weakly suppresses
  6. L19:1068 weakly excites
  7. L19:1513 weakly suppresses
  8. L9:950 weakly suppresses
  9. L7:2404 weakly excites
  10. L23:206 weakly suppresses
  11. L18:2133 weakly suppresses
  12. L23:493 weakly excites
  13. L17:1740 weakly excites
  14. L14:1468 weakly suppresses
  15. L1:2109 weakly suppresses
  16. L17:443 weakly excites
  17. L15:2572 weakly excites
  18. L0:1338 weakly suppresses
  19. L20:335 weakly excites
  20. L4:2909 weakly suppresses
  21. L7:2396 weakly excites
  22. L14:3070 weakly suppresses
  23. L16:752 weakly excites
  24. L21:1060 weakly suppresses

Feeds later layers

  1. L25:2783 suppresses
  2. L26:899 weakly excites
  3. L25:671 weakly excites
  4. L25:2953 weakly suppresses
  5. L25:510 weakly excites
  6. L25:2575 weakly suppresses
  7. L25:2980 weakly excites
  8. L27:434 weakly suppresses
  9. L26:2686 weakly excites
  10. L25:392 weakly suppresses
  11. L25:1645 weakly excites
  12. L26:1070 weakly excites
  13. L25:1583 weakly excites
  14. L25:1635 weakly suppresses
  15. L25:1491 weakly excites
  16. L25:605 weakly excites
  17. L27:2165 weakly excites
  18. L25:2505 weakly suppresses
  19. L26:2570 weakly excites
  20. L26:2449 weakly excites
  21. L25:484 weakly excites
  22. L26:754 weakly suppresses
  23. L27:2604 weakly excites
  24. L27:1533 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.

3 values between -6.17 and -5.561 values between -5.56 and -4.943 values between -4.94 and -4.332 values between -4.33 and -3.711 values between -3.71 and -3.104 values between -3.10 and -2.485 values between -2.48 and -1.8713 values between -1.87 and -1.25116 values between -1.25 and -0.641732 values between -0.64 and -0.021889 values between -0.02 and 0.6097 values between 0.60 and 1.2113 values between 1.21 and 1.838 values between 1.83 and 2.443 values between 2.44 and 3.062 values between 3.06 and 3.671 values between 3.67 and 4.293 values between 4.29 and 4.901 values between 4.90 and 5.523 values between 5.52 and 6.132 values between 6.13 and 6.751 values between 6.75 and 7.361 values between 7.98 and 8.591 values between 11.67 and 12.291 values between 12.90 and 13.521 values between 15.36 and 15.981 values between 17.82 and 18.44offnegativepositiveoften

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. ])))
  3. )').
  4. }})⏎
  5. ))).
  6. }`).
  7. )").
  8. "))))⏎
  9. ))));⏎
  10. )))),

Suppresses

  1. ắn
  2. isky
  3. �인
  4. LEGAL
  5. ника
  6. gings
  7. orage
  8. ности
  9. ник
  10. 痣