Skip to content
mini-lab
Features of prelude-2

Feature #3804

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

Fires on
1.19% of tokens
Mostly in
story
Layer
after block 4

Where it fires

  • story 70%
  • agent 17%
  • chat 10%
  • python 2%
  • worksheet 2%

Makes the model say

. just it that problem met top car

Makes it not say

edit#=**/*.Created trained!")AssertionError Python

Strongest examples

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

  1. story 2.16
    small town, there lived a boy named Tim. Tim had an orange flute that he loved
  2. chat 1.97
    <|user_end|><|assistant_start|>Once upon a time, there was a little cat with soft fur. The cat
  3. story 1.95
    , there was a little girl named Lucy. She had perfect golden hair that she loved to
  4. story 1.93
    there was a gifted little girl named Amy. Amy had a pink dress that she loved very
  5. story 1.92
    asked, "Why are you sad, Bob?" Bob told Willy about the shiny rock and
  6. story 1.89
    favourite toy, a beautiful white
  7. story 1.88
    5.<|assistant_end|><|bos|>One day, a dog named Spot had a messy collar. He did not
  8. story 1.88
    time, there was a little girl named Mia. Mia had a yellow coat that she loved very
  9. story 1.84
    , there was a little dog named Max. Max had a very long tail. One day,
  10. chat 1.8
    about a tree.<|user_end|><|assistant_start|>One day, a large dog with soft fur went for a walk
  11. story 1.76
    he met a new friend, a dog named Max. Max had brown fur and was very kind
  12. story 1.75
    a prince. He was big and strong. He had thick, yellow hair and a kind smile