Skip to content
mini-lab
Features of prelude-2

Feature #1658

Fires on “2” (not labeled yet: read the examples)

Fires on
0.89% of tokens
Mostly in
agent
Layer
after block 4

Where it fires

  • agent 55%
  • python 25%
  • worksheet 8%
  • story 6%
  • chat 5%

Makes the model say

987 lists sp each6A

Makes it not say

Bug * nam For toableOncefrom

Strongest examples

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

  1. python 4.15
    nums)⏎⏎⏎SCORES = [2, 2, 4,
  2. python 3.86
    result + n⏎ return result⏎⏎⏎SCORES = [2, 9, 2,
  3. python 3.81
    2 == 0]⏎⏎⏎SCORES = [6, 2, 5,
  4. python 3.78
    import total, min_item⏎⏎NUMBERS = [9, 6, 3,
  5. python 3.75
    from tools import add_all⏎⏎NUMBERS = [2, 2, 4]
  6. python 3.68
    py⏎from items import head⏎⏎SCORES = [6, 8, 9,
  7. python 3.68
    ⏎ return items[-1]⏎⏎⏎VALUES = [4, 4, 0,
  8. python 3.66
    previous⏎from stats import head⏎⏎SCORES = [9, 6, 2,
  9. python 3.65
    2 == 0]⏎⏎⏎VALUES = [0, 5, 9,
  10. python 3.57
    from listlib import smallest, last⏎⏎NUMBERS = [9, 4, 4,
  11. python 3.53
    py⏎from helpers import first⏎⏎SCORES = [3, 4, 6]
  12. agent 3.53
    , a coding agent harness.<|system_end|><|user_start|>Use [4, 6, 8,