Skip to content
mini-lab
Features of prelude-2

Feature #3992

Fires on “.” (not labeled yet: read the examples)

Fires on
0.44% of tokens
Mostly in
story
Layer
after block 4

Where it fires

  • story 62%
  • agent 23%
  • chat 12%
  • python 2%
  • worksheet 2%

Makes the model say

def best bearFinally = goodbyeitem now

Makes it not say

Oneelore x them gr inside needed

Strongest examples

On text the dictionary never trained on. The underlined token is where it fires most.

  1. story 1.91
    The clown smiled and said, "You're welcome, my friend. I'm glad I could
  2. story 1.9
    Kiki. "It's fun and makes me feel good. Do you want to
  3. story 1.88
    the thief. She said, "Thanks for being my friend. I had a
  4. story 1.81
    , "You are welcome, Tom. You are my friend too. Goodbye." He
  5. story 1.76
    ing.⏎"Thank you, Mia. You are a good friend. A hug is better
  6. story 1.67
    The ball was increasing in size! It was now as big as a house.
  7. story 1.66
    and Ben said.⏎"You're welcome, my sweeties. I love you," mom
  8. story 1.62
    smiles and says, "You're welcome, Tom. I love you. Do you want to
  9. story 1.6
    boy smiled and said, "Sure, I am Danny. I like bricks
  10. story 1.57
    "Of course, I can. I am happy to be your friend. Come, sit with
  11. chat 1.57
    , or some numbers to add?<|assistant_end|><|user_start|>Goodbye!<|user_end|><|assistant_start|>Bye! Come back for
  12. story 1.57
    said, "Thank you, Danny. You are a good friend. Let's play with