Skip to content
mini-lab
Features of prelude-2

Feature #2842

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

Fires on
1.01% of tokens
Mostly in
story
Layer
after block 4

Where it fires

  • story 51%
  • chat 40%
  • agent 7%
  • worksheet 1%
  • python 1%

Makes the model say

songsly " song about stories name words

Makes it not say

sawses adds<|think_start|> needged became>>>

Strongest examples

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

  1. story 8.13
    , Max saw a little bird in the tree. The bird was singing a happy song. Max
  2. story 7.81
    Amy saw a little bird on the tree. The bird was singing a pretty song. Amy wanted
  3. chat 7.13
    In the tree, there was a bird. The bird was singing a pretty poem. The
  4. story 7.02
    ." Tim felt better and started to climb down. The bird sang a happy song to make Tim
  5. story 6.81
    Amy saw a little bird outside the window. The bird was singing a sweet song. Amy called
  6. story 6.77
    The tree was near a deep lake. The bird was happy and sang a pretty song.⏎
  7. story 6.63
    that he was right!⏎The bird was so happy and sang a happy song. To show
  8. story 6.49
    woods near her house. The trees were big, and the birds sang happy songs.⏎As
  9. story 6.41
    Tom. Tom climbed the tree and saw many birds. The birds sang beautiful songs.⏎At
  10. story 6.38
    tree near her house. The bird was blue and yellow and sang very pretty songs. Lila liked
  11. story 6.22
    asked if Max wanted to hear a song.⏎The bird sang a beautiful song for Max.
  12. story 6.07
    were playing, something unexpected happened. The big frog started to sing a beautiful song. Tim was