Skip to content
mini-lab
Log in Sign up
Features of prelude-2

Feature #1712

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

Fires on
0.78% of tokens
Mostly in
worksheet
Layer
after block 4

Where it fires

  • worksheet 40%
  • story 25%
  • python 13%
  • agent 11%
  • chat 11%

Makes the model say

It top the out One it b The

Makes it not say

{ PythonError<|user_start|> counts definedMIT "!"`.

Strongest examples

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

  1. story 2.97
    Bobo was very excited and started to count right away. He counted one, two, three,
  2. story 2.81
    . He took the cloth in his mouth and rubbed it on one of the shoes. Lily
  3. story 2.79
    couldn't contain himself. He plucked one of the ears of
  4. story 2.72
    . Don't look." Lily closed her eyes and counted. "One, two, three,
  5. story 2.68
    to build. They count the blocks as they add them. "One, two, three,
  6. story 2.56
    ⏎As Tim started to climb, he saw a knot on one of the branches. He
  7. story 2.43
    a tree. She wanted to reach the vine so she asked one of the forest animals for
  8. story 2.39
    ten and then seek me." Ben says, "OK. One, two, three..
  9. story 2.3
    er and harder, and the girl saw something drop from one of the trees. She
  10. story 2.29
    unexpected happened. A dog came into the room. The dog took one of the matches and ran
  11. story 2.28
    liked to go on adventures. She was brave and lucky. On one of her adventures she went
  12. story 2.21
    They looked under the bushes and behind the rocks. Finally, one of Tim's friends found