Skip to content
mini-lab
Features of prelude-2

Feature #73

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

Fires on
0.65% of tokens
Mostly in
python
Layer
after block 4

Where it fires

  • python 48%
  • story 32%
  • agent 14%
  • chat 4%
  • worksheet 1%

Makes the model say

who smile's today9 laughance6

Makes it not say

$?" n time." the our when

Strongest examples

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

  1. story 3.78
    Tim was happy to have a new friend.⏎But then
  2. story 3.77
    , "Hello, Sue!" Sue was very happy to have a new friend. They talked and
  3. story 3.76
    and then took a nap. He was so happy to have a new friend and to own a
  4. story 3.75
    to talk and help people. Mia was very happy to have a new friend.⏎One day
  5. story 3.69
    string. Bob was surprised, but he was happy to have a new friend. From that day
  6. story 3.61
    toys back, but he also had something more. He had a new friend, a sister,
  7. story 3.59
    back up to his tree. Joe was so happy to have a new friend and so he sang
  8. story 3.55
    ches them and smiles. She is happy that Sam has found a new friend. She thinks that
  9. story 3.53
    ting me free!" Tim was happy, and now he had a new friend to play with in
  10. story 3.52
    lady was happy too. The cat and the old lady had a new friend. The bird vis
  11. python 3.52
    ⏎ return min(nums)⏎```⏎⏎>>> min_item([3, 9
  12. python 3.51
    _list⏎⏎assert has_item([4, 5, 2, 4