Skip to content
mini-lab
Features of prelude-2

Feature #1960

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

Fires on
0.30% of tokens
Mostly in
story
Layer
after block 4

Where it fires

  • story 85%
  • chat 8%
  • agent 5%
  • python 1%
  • worksheet 0%

Makes the model say

" saidw" I it so Tom

Makes it not say

MITcalculatorsesread Instead VALUES defined WORD

Strongest examples

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

  1. story 5.37
    to increase in size.⏎"Wow!" said Bill. "I didn't
  2. story 5.1
    sorry for his sister.⏎"Poor Sara," he said. "You should have
  3. story 5.09
    just shook her head.⏎"I'm so sorry," she said. "But there's
  4. story 5.08
    little scared.⏎"It's okay, Timmy," his mom said. "I'm right
  5. story 4.9
    ing at him.⏎"I'm sorry, Spot," he said. "You are a
  6. story 4.85
    s.⏎"Hello, little fireman," the fireman says. "Do you want
  7. story 4.82
    , Sam.⏎"Hi, Sam!" said Tim. "Do you want
  8. story 4.72
    , blocking out the light.⏎"Oh no!" Lucy cried. She stepped back and
  9. story 4.55
    ed back at him.⏎"Uh oh," said Peter. "The cow is
  10. story 4.55
    ⏎"You're welcome, little fireman," the fireman says. "You are very
  11. story 4.55
    shield.⏎"Here, these are your toys," she said. "You can play
  12. story 4.52
    sleepy and bored.⏎"Hello, hippo!" Lily said. "Do you want