Feature #2053
Fires on “ sub” (not labeled yet: read the examples)
- Fires on
- 1.87% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 48%
- python 46%
- chat 3%
- story 1%
- worksheet 1%
Makes the model say
([("Added_pathfulultiplecho
Makes it not say
said?n did hole⏎ end on
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
python 5.12) == 6⏎assert sub_one(1
-
python 5.11tail⏎⏎assert tail([8,
-
python 5.0first_letter⏎⏎assert loud("moon") ==
-
python 4.94⏎from strings import final_letter, count_vowels⏎⏎assert final_letter("star
-
python 4.9("Star") == "star"⏎assert lower_case("P
-
python 4.89combine("pizza", "pizza") == "pizza pizza"⏎assert excite("lemon") ==
-
python 4.84tree") == "t"⏎assert join_words("cloud
-
python 4.82abs_value, sum_up_to, divides⏎⏎assert abs_value(-6
-
python 4.82_up_to(12) == 78⏎assert lower_of(1
-
python 4.78combine("robot", "moon") == "robot moon"⏎assert combine("music", "
-
python 4.77⏎from text import repeat, vowels, first_letter⏎⏎assert repeat("tree",
-
python 4.75(x):⏎ return x - 1⏎```⏎⏎>>> sub_one(7