Neuron Explorer

Qwen3-0.6B

the word 'still' mid-sentence

A neuron in layer 4 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

Fires on the word "still" when it sits between a subject and its verb, as in "has still been unable" or "is still evolving." Turning it up did not produce any obvious shared theme in the output.

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

    seizures and that one of the students had previously suffered seizures.⏎But Susan Rafla, the mother of the other child who suffered a seizure, said she has still been unable to "rule in or rule out" the possibility that the portable building

  2. #2 peak

    d'Anticosti and Îles de la Madeleine), Nova Scotia (including present-day N.B. and P.E.I.). Hudson's Bay Co. still administers Rupert's Land. Louisiana is ceded to Spain by France.⏎Following

  3. #3 peak

    manual tools—so that if one kind fails, another can intercede.⏎If everything goes wrong, and your reactor melts down, don't give up. You still have evacuation<|endoftext|>Mohammad H

  4. #4 peak

    the infection, suggests John Jernigan, MD, director of the CDC's office of health care infection prevention, research, and evaluation.⏎"The attack rate is still to be determined, but

  5. #5 peak

    De Stael, a French author, at one time invested in the town lands, on the advice of her friend Gouverneur Morris, while it was still part of Pierrepont.⏎A history in 1894 indicated that

  6. #6 peak

    being gay in some parts of the country. As a result, LGBTQ persons in the United States may still face some challenges not experienced by non-LGBTQ residents, particularly in the Deep

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 am a robot that is not for the public." and then is the example sentence. The first thing is the thing that is not the first thing, and the last thing is the last thing. The first paragraph is about the first

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

    proportional to the change in concentration (C, higher concentration less lower concentration) divided by the thickness of the membrane. At steady state, Fickian diffusion can be approximated by:⏎J = D (C

  2. #2 peak

    -sustaining arc across electrodes placed 5 mm apart. A little math told him that he

  3. #3 peak

    match experimental results found in the literature. These results are for steady-state solutions; a dynamic implementation requires transient modeling as well. This can be achieved by

  4. #4 peak

    to staying up-to-date, boosting your immune system is wellness 101. Better yet

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. L3:2720 weakly suppresses
  2. L0:1751 weakly suppresses
  3. L3:115 weakly suppresses
  4. L0:1885 weakly excites
  5. L2:41 weakly suppresses
  6. L3:2169 weakly excites
  7. L3:2071 weakly excites
  8. L0:2953 weakly excites
  9. L2:2810 weakly excites
  10. L1:2252 weakly suppresses
  11. L3:550 weakly suppresses
  12. L1:1674 weakly excites
  13. L2:803 weakly suppresses
  14. L1:1239 weakly excites
  15. L3:1934 weakly excites
  16. L2:2574 weakly suppresses
  17. L3:1709 weakly suppresses
  18. L1:2408 weakly excites
  19. L2:1888 weakly excites
  20. L2:111 weakly suppresses
  21. L2:1760 weakly suppresses
  22. L1:2917 weakly excites
  23. L2:2624 weakly excites
  24. L1:1440 weakly suppresses

Feeds later layers

  1. L7:1079 excites
  2. L17:2158 excites
  3. L16:2123 excites
  4. L7:1884 excites
  5. L10:696 excites
  6. L13:2890 excites "still" or "as long as" meaning continuing
  7. L11:1853 excites
  8. L9:1699 excites
  9. L18:590 excites
  10. L16:937 suppresses
  11. L13:1147 suppresses
  12. L15:1345 excites
  13. L20:2841 weakly excites
  14. L8:2072 weakly excites
  15. L13:2523 weakly suppresses
  16. L5:358 weakly excites
  17. L15:2779 weakly excites
  18. L9:1892 weakly excites
  19. L26:540 weakly excites
  20. L14:2743 weakly excites
  21. L14:1874 weakly excites
  22. L5:2146 weakly excites
  23. L8:2486 weakly excites
  24. L19:1820 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 -1.18 and -0.7835 values between -0.78 and -0.392531 values between -0.39 and 0.011317 values between 0.01 and 0.4012 values between 0.40 and 0.801 values between 1.19 and 1.591 values between 1.59 and 1.983 values between 2.38 and 2.771 values between 2.77 and 3.171 values between 4.75 and 5.151 values between 10.68 and 11.071 values between 11.47 and 11.861 values between 12.65 and 13.051 values between 13.05 and 13.441 values between 14.23 and 14.63offnegativepositiveoften

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. LAG
  2. `"
  3. `'
  4. pys
  5. AINS
  6. OG
  7. casts
  8. esy
  9. blo
  10. ไ

Suppresses

  1. soever
  2. nex
  3. uç
  4. uthor
  5. SPECIAL
  6. radi
  7. ALLE
  8. rush
  9. 答卷
  10. entral