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

Feature #286

Fires on “6” (not labeled yet: read the examples)

Fires on
0.45% of tokens
Mostly in
python
Layer
after block 4

Where it fires

  • python 71%
  • agent 18%
  • story 7%
  • worksheet 2%
  • chat 2%

Makes the model say

( average://path code around functions divisible

Makes it not say

⏎ also very end started wasAfterfactorial

Strongest examples

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

  1. python 6.59
    _list(SCORES))⏎⏎$ python3 app.py⏎6.0⏎<|bos|>
  2. python 6.21
    , 8, 3, 8, 2])⏎4.8⏎<|bos|>
  3. python 6.13
    ([5, 5, 3, 9])` returns `5.5`.⏎
  4. python 6.04
    _list(SCORES))⏎⏎$ python3 hello.py⏎3.6⏎<|bos|>
  5. python 5.82
    ([9, 3, 1, 5])` returns `4.5`.⏎
  6. python 5.82
    8, 4, 6, 8])` returns `6.2`.⏎
  7. python 5.79
    2⏎```⏎⏎>>> average(7, 4)⏎5.5⏎<|bos|>
  8. python 5.79
    mean_list([3, 8, 7])` returns `6.0`.⏎
  9. python 5.76
    1, 7, 2, 8])` returns `5.2`.⏎
  10. python 5.71
    average_list([8, 9, 8])` returns `8.333
  11. python 5.69
    ⏎$ python3 run.py⏎Hello, Kai!⏎3.0⏎2
  12. python 5.58
    2⏎```⏎⏎>>> mean(8, 9)⏎8.5⏎⏎