Feature #739
Fires on “⏎” (not labeled yet: read the examples)
- Fires on
- 0.66% of tokens
- Mostly in
- python
- Layer
- after block 4
Where it fires
- python 74%
- agent 20%
- story 3%
- worksheet 2%
- chat 1%
Makes the model say
⏎print<|bos|>Addedbed listfrom}
Makes it not say
firstl leftan more magic lot in
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
python 4.864, 6]⏎print(min_item(NUMBERS))⏎print(total([
-
python 4.4COUNT = 6⏎print(larger(9, 7))⏎print(sub_
-
python 4.381, 9]⏎print(max_item(VALUES))⏎print(double_
-
python 4.38(backwards("star"))⏎⏎# items.
-
python 4.33]⏎print(last([5, 3, 7]))⏎print(smallest(
-
python 4.319, 3, 5, 3, 3]))⏎print(head(
-
python 4.241, 4, 9, 0, 2]))⏎print(has_
-
python 4.23WORD = "python"⏎print(count_vowels(WORD))⏎print(final_
-
python 4.235, 6, 6, 2, 6]))⏎print(cubed(
-
python 4.186, 3, 8, 6, 7]))⏎print(combine("
-
python 4.180]⏎print(minus(1, 7))⏎print(evens(
-
python 4.179, 3]⏎print(min_item(VALUES))⏎print(total([