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

Feature #2580

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

Fires on
1.06% of tokens
Mostly in
python
Layer
after block 4

Where it fires

  • python 50%
  • agent 41%
  • story 6%
  • worksheet 2%
  • chat 1%

Makes the model say

emo<|think_end|>alculator]bugsedpytestexpression

Makes it not say

to342 will1 all we

Strongest examples

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

  1. agent 4.81
    : return [n for n in nums if n %
  2. python 4.72
    def only_even(nums):⏎ return [n for n in nums if n %
  3. python 4.6
    def keep_even(nums):⏎ return [n for n in nums if n %
  4. python 4.6
    def only_even(nums):⏎ return [n for n in nums if n %
  5. python 4.57
    def only_even(nums):⏎ return [n for n in nums if n %
  6. agent 4.54
    ⏎2: return [n for n in nums if n %
  7. python 4.54
    def only_even(nums):⏎ return [n for n in nums if n %
  8. python 4.53
    def keep_even(nums):⏎ return [n for n in nums if n %
  9. python 4.52
    def only_even(nums):⏎ return [n for n in nums if n %
  10. python 4.52
    def keep_even(nums):⏎ return [n for n in nums if n %
  11. python 4.49
    python⏎def evens(nums):⏎ return [n for n in nums if n %
  12. python 4.48
    def only_even(nums):⏎ return [n for n in nums if n %