Feature #600
Fires on “positive” (not labeled yet: read the examples)
- Fires on
- 1.09% of tokens
- Mostly in
- python
- Layer
- after block 4
Where it fires
- python 43%
- worksheet 23%
- agent 21%
- chat 11%
- story 3%
Makes the model say
return[-]E()letter⏎ ident
Makes it not say
like with from But a,aring in
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
python 2.76⏎from utils import abs_value, divides⏎⏎assert abs_value(-4) ==
-
python 2.7is_positive, is_odd, halve⏎⏎assert is_positive(-6) ==
-
python 2.68value, sum_up_to, divides⏎⏎assert abs_value(-6) ==
-
python 2.43.py⏎from arith import positive, smaller, product⏎⏎assert positive(-4) ==
-
python 2.39core import is_positive, next_number⏎⏎assert is_positive(4) ==
-
python 2.31py⏎from calc import magnitude, cubed⏎⏎assert magnitude(0) ==
-
python 2.29` returns True if x is positive. For example, `is_positive(4)` returns
-
python 2.28greetings import quiet, greet, join_words⏎⏎assert quiet("TREE")
-
python 2.27+ b⏎⏎$ python3 test_core.py⏎all
-
python 2.27x is greater than zero. For example, `is_positive(4)` returns
-
python 2.23.py⏎from words import quiet, exclaim⏎⏎assert quiet("TIGER")
-
python 2.21assert decrement(11) == 10⏎assert is_positive(3) ==