Feature #3595
Fires on “print” (not labeled yet: read the examples)
- Fires on
- 0.35% of tokens
- Mostly in
- python
- Layer
- after block 4
Where it fires
- python 51%
- agent 22%
- worksheet 11%
- story 10%
- chat 6%
Makes the model say
( test[-Setconfig sub:// don
Makes it not say
+defm all to feelingen started
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
python 3.96= 6⏎print(larger(9, 7))⏎print(sub_one
-
python 3.93divides, decrement⏎⏎print(decrement(9))⏎print(divides(1
-
python 3.92utils import minus, cube⏎⏎print(cube(9))⏎print(minus(
-
python 3.928, 6]⏎print(loud("sun"))⏎print(tail(SCORES
-
python 3.89⏎print(multiply(7, 12))⏎print(sum_up
-
python 3.89⏎print(biggest([7, 2, 0]))⏎print(add_all
-
python 3.88up_to, even⏎⏎print(even(7))⏎print(sum_up
-
python 3.88SIZE = 1⏎print(decrement(SIZE))⏎print(abs_value
-
python 3.88(min_item([4, 9, 8]))⏎print(diffe
-
python 3.88import half, negate⏎⏎print(half(5))⏎print(negate(2
-
python 3.87, 3]⏎print(min_item(VALUES))⏎print(total([4
-
python 3.86⏎WORD = "robot"⏎print(cubed(6))⏎print(initial(WORD