Neuron Explorer

Qwen3-0.6B

'of' in 'one of the' phrases

A neuron in layer 17 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 'of' right after 'one', as in 'one of the best' or 'one of my patients'. When it was turned up, the model's replies drifted into wordy, rambling text, which doesn't clearly reflect the trigger.

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

    should consume to treat a specific medical condition<|endoftext|>A species on the brink – Freshwater Pearl Mussel⏎They can live for over a century, they have one of the most bizarre life-cycles of any species you’re likely to find and they

  2. #2 peak

    .D. Anderson Cancer Center in Houston. “I’ve always felt the same way, and I tell my patients the same thing.”⏎“Just last week, one of my melanoma patients came in with a sunburn all over their back, and

  3. #3 peak

    this article should be taken as medical advice. Please talk to your doctor before using any of the herbs and/or remedies mentioned in this article. Yarrow is one of the oldest medicinal herbs, dating back thousands of years and used for various ailments.

  4. #4 peak

    Agatha Christie’s mysteries. Christie often portrayed the detective as referring to himself in the third person, as a way of depicting his self-grandeur. In one of her books, another character asks him about it:⏎- Dr. Lutz:

  5. #5 peak

    and consequently the subjects included in<|endoftext|>Authored by Gayathri Vasudevan, Sambhav by LabourNet⏎Did you know that India has one of the best recycling rates in the world? 70% of all PET bottles

  6. #6 peak

    fish abundance and distribution trends between locations. Armed with the knowledge of where<|endoftext|>Edited by William Byron Forbush This is a book that will never die — one of the great English classics. . .

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: that of how, in and by, of time, a person can see and by time, of a person in new and by, of new and by, of new, person, in new and by, of new and by, of

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

    -hour recall of what they had consumed. Only 25 percent reported eating any meat, which historically was a diet staple, and many of them had only had one meal in 24 hours. Some had not eaten at all.⏎The Central

  2. #2 peak

    father.⏎People who inherit the transmuted hemoglobin gene from both biological parents have sickle cell anemia. People who inherit the transmuted gene from one parent have the sickle cell trait.⏎What Raises One’s Risk for Sickle

  3. #3 peak

    S U.⏎No direct anagrams for focus found in this word list.⏎List shorter words within focus, sorted by length⏎All words formed from focus by changing one letter⏎Browse words starting with focus by next letter<|endoftext|>I was listening to Brooke

  4. #4 peak

    as hypomania. Hypomania is one of the most distinguishing bipolar type 2 signs. In order to be classified as bipolar type II, at least one hypomanic episode must occur and must last at least four days.⏎In severe

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. L16:1012 weakly excites 'one' before 'of' in 'is one of the'
  2. L16:551 weakly excites
  3. L5:133 weakly excites
  4. L3:283 weakly excites
  5. L2:2828 weakly suppresses
  6. L0:1 weakly suppresses
  7. L13:1720 weakly suppresses
  8. L10:1575 weakly excites
  9. L1:2355 weakly excites
  10. L2:126 weakly excites
  11. L7:806 weakly excites
  12. L14:17 weakly suppresses
  13. L10:2627 weakly suppresses
  14. L5:2272 weakly excites
  15. L15:280 weakly excites
  16. L16:222 weakly suppresses
  17. L5:1713 weakly suppresses
  18. L14:1764 weakly suppresses
  19. L16:143 weakly excites
  20. L6:1782 weakly suppresses 'of' after a category noun
  21. L1:818 weakly excites
  22. L0:2785 weakly suppresses
  23. L15:1454 weakly excites
  24. L9:2817 weakly suppresses

Feeds later layers

  1. L20:807 excites
  2. L24:434 excites
  3. L24:489 suppresses
  4. L19:654 weakly suppresses
  5. L26:441 weakly suppresses
  6. L23:79 weakly excites
  7. L24:360 weakly suppresses
  8. L18:520 weakly excites
  9. L19:2593 weakly excites
  10. L19:1570 weakly suppresses
  11. L21:913 weakly excites
  12. L19:2358 weakly excites "any of" before "these" or "the"
  13. L25:390 weakly suppresses
  14. L22:260 weakly suppresses
  15. L18:1433 weakly excites
  16. L27:1260 weakly excites
  17. L25:284 weakly excites
  18. L21:918 weakly suppresses
  19. L22:211 weakly suppresses
  20. L19:1695 weakly suppresses
  21. L25:320 weakly suppresses
  22. L20:602 weakly suppresses
  23. L21:392 weakly excites
  24. L19:959 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.

13 values between -2.93 and -1.043848 values between -1.04 and 0.8524 values between 0.85 and 2.745 values between 2.74 and 4.632 values between 4.63 and 6.523 values between 6.52 and 8.421 values between 8.42 and 10.311 values between 10.31 and 12.201 values between 12.20 and 14.091 values between 15.98 and 17.871 values between 27.32 and 29.211 values between 29.21 and 31.101 values between 31.10 and 32.992 values between 32.99 and 34.881 values between 40.55 and 42.441 values between 42.44 and 44.331 values between 44.33 and 46.221 values between 70.80 and 72.69offnegativepositiveoften

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. ials
  2. ités
  3. eties
  4. align
  5. ologies
  6. them
  7. аст
  8. them
  9. achers
  10. ilities

Suppresses

  1. 网首页
  2. 一个小
  3. ONGL
  4. chù
  5. sebuah
  6. группа
  7. carpet
  8. awy
  9. 침
  10. -dat