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

Feature #3295

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

Fires on
0.48% of tokens
Mostly in
agent
Layer
after block 4

Where it fires

  • agent 44%
  • python 33%
  • chat 14%
  • worksheet 7%
  • story 2%

Makes the model say

_ code plus course // toolsuteoldString

Makes it not say

#em! When put ord initial lot

Strongest examples

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

  1. python 5.64
    ⏎ for i in range(n
  2. agent 5.39
    newString= for i in range(1
  3. python 4.97
    _up_to(n):⏎ acc = 0⏎ for i in range(n
  4. python 4.89
    = 1⏎ for n in nums:⏎
  5. agent 4.63
    * 2 for n in nums]`. The
  6. agent 4.61
    n * 2 for n in nums]⏎⏎
  7. python 4.6
    def count_vowels(s):⏎ k = 0⏎ for c in s:⏎
  8. python 4.57
    ⏎⏎⏎def total(nums):⏎ result = 0⏎ for n in nums:⏎
  9. python 4.56
    def count_vowels(s):⏎ k = 0⏎ for c in s:⏎
  10. python 4.56
    _up_to(n):⏎ acc = 0⏎ for i in range(n
  11. python 4.51
    double_all(nums):⏎ return [n * 2 for n in nums]⏎⏎
  12. agent 4.49
    return [n * 2 for n in nums]`. The