Feature #1048
Fires on “ ” (not labeled yet: read the examples)
- Fires on
- 1.48% of tokens
- Mostly in
- python
- Layer
- after block 4
Where it fires
- python 32%
- worksheet 31%
- agent 22%
- chat 11%
- story 4%
Makes the model say
b both big let sure shiny sh I
Makes it not say
Renamed trained!")replaceAll PythonEditedpatternAssertionError
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
python 4.01first([7, 8, 7, 1]) == 7⏎assert increment
-
python 3.7⏎⏎assert first([8, 2, 7]) == 8⏎assert first
-
python 3.41, 7, 5, 8, 4]) == 5⏎assert average
-
python 3.4numbers. For example, `larger(6, 9)` returns `9`.⏎⏎
-
python 3.39biggest([2, 8, 1, 3]) == 8⏎print("
-
python 3.33first([6, 1, 7, 3]) == 6⏎assert first
-
python 3.3biggest([0, 8, 7, 3]) == 8⏎print("
-
python 3.27head([7, 5, 8, 8])` returns `7`.⏎<|bos|>
-
python 3.25biggest([5, 9, 6, 6]) == 9⏎print("
-
python 3.24biggest([3, 6, 2, 8]) == 8⏎assert doubled
-
python 3.21first([4, 8, 7, 5])` returns `4`.⏎<|bos|>
-
python 3.18biggest([0, 5, 4, 6]) == 6⏎print("