Neuron Explorer

Qwen3-0.6B

word right before a term's name

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

This neuron lights up on the word that introduces a name, like 'called', 'known as' or 'named', right before the thing being named. When it was turned up, the text often drifted toward phrases like 'is called' or 'the system is called'.

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

    daily functioning. This condition, known as mild cognitive impairment, progresses to Alzheimer’s in roughly 10% to

  2. #2 peak

    its axis. In such movement particle has charge, produce electric waves ( waves property of particle) and its speed ( frequency) is : c>1. We call such particle - ? electron? and its energy is: E=h*f.⏎In

  3. #3 peak

    It’s called a phone booth.⏎Phone booths used to be everywhere, providing an office for agent

  4. #4 peak

    Peritoneal Cancer: Causes, Symptoms, Diagnosis, …⏎(2 days ago) The inside of your abdomen has a slick, transparent lining. It’s called the peritoneum. When

  5. #5 peak

    of you who don’t know what a genetic algorithm is, it is basically an optimisation method that mimics evolution. You have a set of candidate solutions (called a population). The best solutions get combined with each other (mating) and

  6. #6 peak

    , deer have food they avoid, and<|endoftext|>Seven miles to the south west of Sheffield there's a piece of wild moorland surrounded by roads. It's called Big Moor. I estimate that if you circumnavigated it on foot by

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: 20% of the month, 10% of the year, 30% of the year. This month, 10% of the year, 10% of the year, 10% 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

    For backgrounds, remember yin and yang. If you use a light background, use dark or black colour for text. If you have a dark background, use white or light colour for text. The text needs to stand out from the background.⏎

  2. #2 peak

    .⏎In the examples shown, we used stickers (rooms and stars) from our In the house sticker pack. We drew the other symbols onto blank stickers.⏎#2 Outdoor I-Spy⏎Use your daily walk to play I- Spy and spot

  3. #3 peak

    1785-87 (circa)⏎Publication: Troisieme Voyage De Cook, ou Voyage a l'Ocean Pacifique…⏎Color: Black & White⏎The copper engravings from

  4. #4 peak

    . It produces dense clusters of hanging tubular flowers, the colour varies from white, pink, red, green to yellow. The corolla is a partially inflated

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. L3:2662 weakly suppresses
  2. L2:75 weakly suppresses
  3. L3:272 weakly excites
  4. L3:476 weakly excites
  5. L3:1001 weakly suppresses
  6. L3:967 weakly excites
  7. L0:1734 weakly excites
  8. L2:2141 weakly excites
  9. L3:1773 weakly suppresses
  10. L0:2837 weakly suppresses
  11. L3:59 weakly excites
  12. L2:36 weakly excites
  13. L1:87 weakly suppresses
  14. L0:3029 weakly excites
  15. L0:2132 weakly excites
  16. L0:1679 weakly excites
  17. L1:691 weakly suppresses
  18. L2:2201 weakly suppresses
  19. L2:1275 weakly suppresses
  20. L3:1757 weakly suppresses
  21. L2:111 weakly suppresses
  22. L1:1689 weakly excites
  23. L3:2893 weakly suppresses
  24. L2:2630 weakly excites

Feeds later layers

  1. L5:944 excites
  2. L6:1197 excites
  3. L8:1224 excites
  4. L12:2252 excites
  5. L5:2715 suppresses
  6. L11:1231 weakly excites
  7. L5:1317 weakly excites 'a' right after 'being'
  8. L22:2344 weakly suppresses number after 'account for' or 'over'
  9. L7:384 weakly suppresses
  10. L5:647 weakly excites
  11. L12:2902 weakly excites
  12. L10:972 weakly suppresses
  13. L5:1183 weakly excites
  14. L15:1977 weakly excites
  15. L21:945 weakly excites
  16. L6:1258 weakly excites "mean by" in questions and definitions
  17. L18:1329 weakly excites
  18. L8:1348 weakly suppresses
  19. L17:2516 weakly excites
  20. L8:2122 weakly excites hand and body action verbs in instructions
  21. L6:662 weakly excites
  22. L12:74 weakly excites
  23. L7:1128 weakly suppresses
  24. L18:2718 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 -1.64 and -1.551 values between -1.21 and -1.131 values between -0.78 and -0.701 values between -0.70 and -0.611 values between -0.61 and -0.533 values between -0.53 and -0.4410 values between -0.44 and -0.3642 values between -0.36 and -0.27188 values between -0.27 and -0.19414 values between -0.19 and -0.10936 values between -0.10 and -0.021283 values between -0.02 and 0.07672 values between 0.07 and 0.16255 values between 0.16 and 0.2470 values between 0.24 and 0.3324 values between 0.33 and 0.412 values between 0.41 and 0.501 values between 0.58 and 0.671 values between 0.67 and 0.751 values between 0.93 and 1.011 values between 1.69 and 1.78offnegativepositiveoften

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. 叫做
  2. called
  3. named
  4. 叫
  5. 简称
  6. �
  7. 称号
  8. 名
  9. name
  10. 名字

Suppresses

  1. Wilkinson
  2. enville
  3. spilled
  4. ügen
  5. vore
  6. prefs
  7. ordan
  8. ologne
  9. opacity
  10. :normal