Feature #447
Fires on “print” (not labeled yet: read the examples)
- Fires on
- 0.41% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 46%
- python 45%
- story 4%
- chat 3%
- worksheet 1%
Makes the model say
(addyp addfromplusn
Makes it not say
todayevensreturn decided evens⏎ wishedeven
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
python 5.86⏎# script.py⏎from math_utils import product⏎⏎print(product(1
-
python 5.81arith import sum_to, negate⏎⏎SIZE = 5⏎print(sum_to
-
python 5.68math_utils import difference⏎from helpers import min_item⏎⏎print(min_item
-
python 5.68]⏎⏎# main.py⏎from textlib import loud⏎⏎print(loud("cat
-
python 5.66.py⏎from lib import fact⏎⏎LIMIT = 2⏎print(fact(LIMIT
-
python 5.62⏎# main.py⏎from calc import positive, product⏎⏎print(product(7
-
python 5.62.py⏎from ops import double⏎⏎LIMIT = 3⏎print(double(LIMIT
-
python 5.6<|bos|># main.py⏎from words import count_vowels⏎⏎print(count_vowels
-
python 5.59⏎# script.py⏎from lib import max_item⏎⏎print(max_item
-
python 5.59"!"⏎⏎# main.py⏎from lib import hello⏎⏎print(hello("E
-
python 5.58main.py⏎from strings import join_words, welcome⏎⏎print(welcome("Mia
-
python 5.56import sum_up_to⏎⏎LIMIT = 11⏎print(sum_up