Skip to content
mini-lab
Features of prelude-2

Feature #211

Fires on “0” (not labeled yet: read the examples)

Fires on
0.68% of tokens
Mostly in
story
Layer
after block 4

Where it fires

  • story 59%
  • python 25%
  • chat 8%
  • agent 7%
  • worksheet 2%

Makes the model say

])])` jobOnce by shinyayonfig

Makes it not say

" moment⏎ist havebiggest two

Strongest examples

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

  1. python 2.19
    ⏎⏎>>> total([2, 4, 1, 0, 4])
  2. python 2.16
    >>> sum_list([7, 4, 6, 0, 6])
  3. story 2.14
    from the police. But the police were fast too. The car got caught by the police.
  4. python 2.14
    >>> mean_list([5, 7, 4, 3])⏎4.
  5. python 2.13
    assert sum_list([7, 0, 4, 0, 5])
  6. python 2.01
    assert add_all([5, 2, 6, 0, 2])
  7. python 1.98
    >>> add_all([8, 0, 0, 6])⏎14
  8. python 1.93
    assert keep_even([0, 8, 7, 6, 9])
  9. story 1.88
    and turned into a bright and colourful jelly. His jelly was saved! Mikey
  10. python 1.87
    >>> how_many([5, 4, 4, 2])⏎4⏎
  11. python 1.86
    >>> average_list([0, 4, 6, 4])⏎3.
  12. agent 1.85
    def tail<|tool_call_end|><|assistant_end|><|tool_start|>