Skip to content
mini-lab
Features of prelude-2

Feature #2689

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

Fires on
0.90% of tokens
Mostly in
story
Layer
after block 4

Where it fires

  • story 80%
  • chat 12%
  • agent 5%
  • worksheet 3%
  • python 1%

Makes the model say

no. sad,?" clappedWow now

Makes it not say

Please editOnceassertllTell_*.

Strongest examples

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

  1. story 3.32
    zed them too hard.⏎The sunglasses broke into pieces. Tim was sad because his new sunglasses
  2. story 3.32
    the rope. The rope broke, and Tim fell down. Tim was sad and hurt. He
  3. story 3.25
    her toy chest broke and all her toys fell out. Amy was sad.⏎Amy's
  4. story 3.24
    the round ball got stuck in the tree. Tim and his friends felt sad. A kind man
  5. chat 3.12
    cake off the table. The cake fell and was broken. Tim was sad, but the birds
  6. story 3.09
    in a pond and started to float. Tim and the bug were sad, but then they
  7. story 3.09
    A big wind came and blew the tie away. Tom and Sam were sad. They looked and
  8. story 3.06
    Tom's ice-cream fell on the ground. Tom was sad and cried. Lily
  9. story 3.03
    knocked down the train station. The blocks fell everywhere. Tim felt sad, but Sarah said
  10. story 3.03
    and blew the nest away. Bessie and the little bird were sad. But Bess
  11. chat 3.02
    wet grass. His toy gun fell out of his hands. Tim was sad, but his mom
  12. story 3.01
    ing. But the cup fell off her head and broke. Lily was sad and Ben was worried