Feature #793
Fires on “1” (not labeled yet: read the examples)
- Fires on
- 0.62% of tokens
- Mostly in
- python
- Layer
- after block 4
Where it fires
- python 46%
- agent 38%
- chat 9%
- worksheet 4%
- story 3%
Makes the model say
)py+20): +])
Makes it not say
But ButOh why momentThis⏎⏎ me
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
python 5.01_up_to(1) == 1⏎assert positive(12) == True
-
python 4.86) == 6⏎assert sub_one(1) == 0
-
python 4.76minus(4, 6) == -2⏎assert half(1) == 0
-
python 4.47⏎ return -x⏎ return x⏎```⏎⏎>>> absolute(11)⏎1
-
python 4.46, 5) == 5⏎assert product(5, 10) ==
-
python 4.43sub_one, square, mean⏎⏎assert sub_one(12) ==
-
python 4.42.py⏎from helpers import absolute, doubled⏎⏎assert absolute(11) ==
-
python 4.410⏎assert opposite(4) == -4⏎assert opposite(1) == -1
-
python 4.38smaller(5, 3) == 3⏎assert twice(11) ==
-
python 4.38cubed` returns the cube of x. For example, `cubed(1)` returns `1
-
python 4.27⏎ return -x⏎ return x⏎```⏎⏎>>> absolute(10)⏎1
-
python 4.25True⏎assert decrement(9) == 8⏎assert decrement(1) == 0