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

Feature #215

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

Fires on
1.21% of tokens
Mostly in
story
Layer
after block 4

Where it fires

  • story 48%
  • agent 36%
  • chat 6%
  • python 5%
  • worksheet 4%

Makes the model say

everyone that("led from who playing of

Makes it not say

oldString uponnewString⏎ How WhyvSet

Strongest examples

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

  1. agent 1.88
    can list, read, run
  2. python 1.79
    " " + b⏎```⏎⏎>>> join_words("cloud", "cat")⏎
  3. story 1.78
    drink water every day. He knew
  4. story 1.77
    Spot, went for a walk. They saw a big tree with a sign that said "V
  5. python 1.73
    tree") == "t"⏎assert join_words("cloud", "sun") ==
  6. python 1.67
    " " + b⏎```⏎⏎>>> join_words("hello", "code")⏎
  7. story 1.66
    from the storm. But then, they saw a sign that said, "
  8. story 1.66
    can do special things."⏎John
  9. story 1.65
    I said you can eat it. But lemons are sour, not sweet. You have
  10. python 1.65
    ated by a space. For example, `join_words("apple", "sun")` returns
  11. story 1.65
    . I tried to tell you. It's a razor, not a comb. You
  12. story 1.64
    , Amy and Tom found a store that sold balls. They saw a sign that said "P