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

Feature #1567

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

Fires on
0.49% of tokens
Mostly in
agent
Layer
after block 4

Where it fires

  • agent 42%
  • python 30%
  • story 22%
  • chat 3%
  • worksheet 2%

Makes the model say

y dog is was ball cat( and

Makes it not say

otherched>=**/ wish fixmodule sister

Strongest examples

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

  1. python 2.84
    larger, sub_one⏎⏎COUNT = 6⏎print(larger(9,
  2. python 2.82
    demo.py⏎from mathlib import power, cube⏎⏎print(power(2,
  3. python 2.8
    main.py⏎from calc import positive, product⏎⏎print(product(7,
  4. python 2.71
    script.py⏎from math_utils import product⏎⏎print(product(10,
  5. python 2.68
    py⏎from number_utils import minus, cube⏎⏎print(cube(9))⏎
  6. python 2.68
    return a * b⏎⏎⏎LIMIT = 6⏎print(multiply(4,
  7. python 2.67
    previous, add_one⏎⏎LIMIT = 3⏎print(previous(LIMIT))⏎
  8. python 2.6
    hello.py⏎from mathlib import larger, difference⏎⏎print(larger(2,
  9. python 2.58
    arith import larger⏎from tools import add_all⏎⏎print(larger(8,
  10. python 2.56
    (nums):⏎ return min(nums)⏎⏎⏎print(multiply(5,
  11. python 2.49
    ⏎# main.py⏎from helpers import product⏎⏎print(product(2,
  12. python 2.48
    sum_to, negate⏎⏎SIZE = 5⏎print(sum_to(2