Skip to content
mini-lab
Features of prelude-2

Feature #86

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

Fires on
0.59% of tokens
Mostly in
worksheet
Layer
after block 4

Where it fires

  • worksheet 55%
  • python 17%
  • chat 13%
  • agent 12%
  • story 3%

Makes the model say

many3( were other2 with promised

Makes it not say

gw why])`OneOncemd turns

Strongest examples

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

  1. worksheet 4.4
    645 + 9 = 654.⏎If you add 1 and
  2. worksheet 4.39
    3 is 568.⏎If you add 15
  3. worksheet 4.33
    64 = 115585.⏎If you add 11
  4. worksheet 4.32
    56 + 8 = 7964.⏎If you add 9 and
  5. worksheet 4.32
    and 14 is 7749.⏎If you add 63
  6. worksheet 4.29
    ⏎If you add 47
  7. worksheet 4.27
    6 + 663 = 669.⏎If you add 74
  8. worksheet 4.07
    If you add 24
  9. worksheet 3.96
    7644 = 37726.⏎If you add 92
  10. worksheet 3.95
    8603 = 8676.⏎If you add 44
  11. worksheet 3.93
    47 + 72 = 119.⏎If you add 66
  12. worksheet 3.91
    0 + 787 = 797.⏎If you add 81