Skip to content
mini-lab
Features of prelude-2

Feature #881

Fires on “3” (not labeled yet: read the examples)

Fires on
0.83% of tokens
Mostly in
python
Layer
after block 4

Where it fires

  • python 55%
  • agent 32%
  • worksheet 6%
  • story 5%
  • chat 2%

Makes the model say

6owedonlylist different function agreed7

Makes it not say

plan ear example sawt night idea One

Strongest examples

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

  1. agent 4.49
    , 8, 7]) == 6.66666666
  2. python 4.38
    , 5, 5]) == 3.66666666
  3. python 4.28
    9, 9, 4])⏎7.33333333
  4. python 4.25
    ⏎assert total([4, 8, 4]) == 16⏎assert min
  5. python 4.24
    , 0, 9]) == 4.33333333
  6. python 4.11
    main.py⏎[8]⏎4.333333333
  7. python 4.06
    , 6, 0])` returns `2.66666666
  8. python 3.97
    ([1, 7, 7, 4]) == 19⏎print("
  9. python 3.97
    , 1, 3]) == 3.66666666
  10. python 3.89
    main.py⏎2⏎8.33333333333
  11. python 3.85
    , 9, 8])` returns `8.33333333
  12. python 3.73
    sum_list([9, 7, 3]) == 19⏎assert join