Feature #566
Fires on “def” (not labeled yet: read the examples)
- Fires on
- 0.69% of tokens
- Mostly in
- python
- Layer
- after block 4
Where it fires
- python 68%
- agent 29%
- story 2%
- chat 0%
- worksheet 0%
Makes the model say
so evendef return7 is1 p
Makes it not say
<|arg|>expressionOUNTpositive):>divEXT
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
python 7.98but it has `return a * a`.⏎⏎```python⏎def product(a,
-
python 7.78numbers, but it has `return a`.⏎⏎```python⏎def max_of(
-
python 7.65numbers, but it has `return a`.⏎⏎```python⏎def max_of(
-
python 7.47but it has `if a < b:`.⏎⏎```python⏎def max_of(
-
python 7.45x)` returns the absolute value of x.⏎⏎```python⏎def abs_value(
-
agent 7.36can I find biggest?<|user_end|><|assistant_start|><|tool_call_start|>grep<|arg|>pattern=def biggest<|tool_call_end|><|assistant_end|><|tool_start|>
-
python 7.34)` returns the average of a and b.⏎⏎```python⏎def average(a,
-
python 7.31(s)` returns s in lower case.⏎⏎```python⏎def whisper(s):
-
agent 7.29find join_words?<|user_end|><|assistant_start|><|tool_call_start|>grep<|arg|>pattern=def join_words<|tool_call_end|>
-
python 7.23has `return a + "" + b`.⏎⏎```python⏎def join_words(
-
python 7.19has `return a + "" + b`.⏎⏎```python⏎def combine(a,
-
python 7.19return sum(nums) / 2`.⏎⏎```python⏎def mean_list(