Neuron Explorer

Qwen3-0.6B

'that' starting a new clause

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' when it is used like 'this' or 'that' to refer back to something just said, as in 'and that percentage' or 'of course that isn't always the case'. When it was turned up, the model's text drifted into rambling and repetition rather than a clear theme.

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

    é’s population is 2.5 million. The urbanization rate is currently 3.3%; 58% of the country is urbanized and that percentage continues to grow annually.⏎Cameroon’s population is made up of its indigenous

  2. #2 peak

    <|endoftext|>Homeowners across a large swath of Pennsylvania are reporting sightings of spotted lanternflies hatching from egg masses.⏎“Egg-hatching season is here, and that has some people on edge,” said Emelie Swackhamer, a

  3. #3 peak

    . You can communicate a thought, but your thought may not be understood. In some cases, your thought may not even reach the proper target. Thats why it pays to ask questions to make certain that people understand what you are saying

  4. #4 peak

    good listener.⏎Over a period of 240 years there has been 44 presidents and the U.S has had both good times and bad times and that all depended on who was running the country and what choices they

  5. #5 peak

    stakes and adrenaline during a game. (Or a whimsical ‘why not’). But of course that isn’t always the case.⏎A total of 64% of Australian adults

  6. #6 peak

    00. But of course that’s not what you’ll actually receive.⏎Your employer is required to withhold federal 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: What is the value of the parameter in the parameterization of the curve $ \frac{1}{0} $ when $ \text{0} $, $ \text{1} $, $ \text{1} $,

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

    2. Every day the boss enters the office⏎at half past nine⏎3. My friend will ring me up at eight o’clock⏎4. Where was Kate going when you met her yesterday?⏎does your brother work?⏎6. What

  2. #2 peak

    God Himself standing over him in judgment. “Where were you hurt?” the general asked in his high-pitched voice (Patton sounded

  3. #3 peak

    expelled the only red state in the national organizaton" -- but<|endoftext|>Raising Our Children to Be Gatekeepers⏎And the LORD said unto Cain, Where is Abel thy brother? And he said, I know not: Am I my brother

  4. #4 peak

    )<|endoftext|>High: 77°F ~ Low: 59°F⏎Monday, Sep. 15, 2014⏎"Where's The Beef"Posted Friday, October 9, 2009,

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. L3:700 excites
  2. L3:828 suppresses
  3. L3:487 weakly suppresses
  4. L3:854 weakly suppresses
  5. L1:74 weakly excites
  6. L1:1273 weakly excites
  7. L1:2689 weakly suppresses
  8. L2:2681 weakly suppresses
  9. L1:1053 weakly suppresses
  10. L3:246 weakly excites
  11. L0:593 weakly excites
  12. L2:2557 weakly suppresses
  13. L3:762 weakly suppresses
  14. L3:155 weakly excites
  15. L3:745 weakly excites
  16. L1:1169 weakly excites
  17. L3:189 weakly excites
  18. L0:2338 weakly suppresses
  19. L3:569 weakly excites unclear: mid-phrase function words
  20. L2:1527 weakly excites
  21. L3:1068 weakly excites "that" pointing back to a noun
  22. L1:29 weakly suppresses
  23. L2:2830 weakly excites
  24. L2:2137 weakly suppresses

Feeds later layers

  1. L10:965 excites the word 'that' introducing a clause
  2. L17:1821 weakly excites
  3. L15:2787 weakly excites
  4. L20:586 weakly excites
  5. L15:810 weakly suppresses
  6. L20:2799 weakly suppresses
  7. L14:1540 weakly excites
  8. L17:1851 weakly excites
  9. L24:54 weakly excites
  10. L11:1110 weakly excites
  11. L26:2486 weakly suppresses
  12. L15:470 weakly excites the word 'this' after a preposition
  13. L22:1365 weakly suppresses
  14. L17:2680 weakly excites
  15. L10:1770 weakly excites
  16. L24:1345 weakly excites
  17. L11:2788 weakly suppresses 'side' in 'side effects'
  18. L19:1935 weakly excites
  19. L9:917 weakly excites
  20. L14:84 weakly suppresses
  21. L9:2011 weakly excites
  22. L18:789 weakly excites
  23. L12:1812 weakly excites
  24. L19:1004 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 -1.46 and -1.254 values between -1.04 and -0.8312 values between -0.83 and -0.6221 values between -0.62 and -0.41194 values between -0.41 and -0.202017 values between -0.20 and 0.011501 values between 0.01 and 0.22118 values between 0.22 and 0.4311 values between 0.43 and 0.647 values between 0.64 and 0.855 values between 0.85 and 1.062 values between 1.06 and 1.273 values between 1.27 and 1.481 values between 1.48 and 1.691 values between 1.69 and 1.902 values between 1.90 and 2.111 values between 2.11 and 2.321 values between 2.32 and 2.531 values between 2.95 and 3.162 values between 3.16 and 3.371 values between 4.20 and 4.412 values between 6.72 and 6.93offnegativepositiveoften

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. PGA
  2. followed
  3. asc
  4. uada
  5. eway
  6. register
  7. alta
  8. 唬
  9. registro
  10. asthan

Suppresses

  1. kräfte
  2. ện
  3. וץ
  4. äl
  5. quartered
  6. -floor
  7. vọng
  8. 揆
  9. 韧
  10. ?q