Feature #3649
Fires on “python” (not labeled yet: read the examples)
- Fires on
- 0.84% of tokens
- Mostly in
- python
- Layer
- after block 4
Where it fires
- python 46%
- worksheet 32%
- chat 11%
- agent 10%
- story 1%
Makes the model say
<|assistant_end|> PythoncommandRenamedreplaceAll't function generator
Makes it not say
in, areE she andOne
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
python 2.17x + x` with `return x * x`.⏎⏎```python⏎def squared(
-
python 2.161` with `return x + 1`.⏎⏎```python⏎def increment(
-
python 2.13: replace `return a` with `return b`.⏎⏎```python⏎def larger(
-
python 2.11` with `return x + 1`.⏎⏎```python⏎def next_
-
python 2.08, 11) == 11⏎print("all tests passed")⏎⏎#
-
python 2.07` with `return a % b == 0`.⏎⏎```python⏎def divides(
-
python 2.06x` with `return x * x * x`.⏎⏎```python⏎def cube(
-
python 2.06x + x` with `return x * x`.⏎⏎```python⏎def square(
-
python 2.055` with `return x > 0`.⏎⏎```python⏎def positive(
-
python 2.04not in items` with `return x in items`.⏎⏎```python⏎def has_
-
python 2.04not in items` with `return x in items`.⏎⏎```python⏎def has_
-
python 2.03: replace `return a` with `return b`.⏎⏎```python⏎def max_