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

Feature #1189

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

Fires on
0.42% of tokens
Mostly in
agent
Layer
after block 4

Where it fires

  • agent 37%
  • python 37%
  • chat 18%
  • story 5%
  • worksheet 4%

Makes the model say

to]) "ient helppy test its

Makes it not say

." shapeselrite`,)`. h days

Strongest examples

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

  1. python 2.7
    assert smallest([7, 5, 9, 6, 3])
  2. python 2.56
    >>> mean_list([5, 7, 4, 3])⏎4.
  3. python 2.55
    assert sum_list([7, 0, 4, 0, 5])
  4. python 2.51
    assert keep_even([7, 7, 5, 3, 1])
  5. worksheet 2.5
    804 = 1791⏎Could you tell me what 44
  6. worksheet 2.47
    0 = 55310.⏎Could you tell me what 22
  7. worksheet 2.47
    ⏎7 plus 5 equals 12.⏎Could you tell me what 60
  8. worksheet 2.45
    . Now Tim has 326 stars.⏎Could you tell me what 5 +
  9. python 2.45
    >>> how_many([5, 4, 4, 2])⏎4⏎
  10. python 2.44
    ⏎⏎assert last([8, 8, 8, 4, 6])
  11. worksheet 2.43
    + 0 = 9262.⏎Could you tell me what 99
  12. worksheet 2.42
    7 + 6 = 683.⏎Could you tell me what 53