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

Feature #3469

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

Fires on
0.88% of tokens
Mostly in
story
Layer
after block 4

Where it fires

  • story 54%
  • agent 38%
  • chat 7%
  • python 1%
  • worksheet 1%

Makes the model say

it the this them from about every which

Makes it not say

functions<|assistant_end|><|tool_start|>ood<|tool_call_end|><|system_end|><|system_start|>newString

Strongest examples

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

  1. story 3.5
    sandwiches and went outside. Tim was happy to eat outside and watch the jets.
  2. story 3.46
    Tim liked to watch the fish swim.
  3. story 3.4
    . The pie smelled great.⏎Lucy and her mom sat down to eat the pie. They
  4. story 3.38
    . Lucy was a bit lazy, so she sat on a bench to eat her lollipop.⏎
  5. story 3.38
    red book. Her aunt smiled and said, "Yes, let's read it."⏎As
  6. chat 3.34
    very happy. She bought the novel and went home to read it.⏎When
  7. chat 3.29
    , he sat on a bench and watched the cake.⏎
  8. story 3.27
    cake, and ice cream. Tom said, "Max, let's eat some of these sweet
  9. story 3.26
    . It was a big, soft cake.⏎Mark wanted to eat the cake, but
  10. story 3.25
    's mom put the soup on the table. Sue was so excited to eat the soup. But
  11. chat 3.22
    Oh, these apples look so tasty," shouted Sam. "Can I eat one?"⏎His
  12. story 3.18
    very original. She was so happy with her pie and wanted to share it with her friends