Feature #3383
Fires on “utils” (not labeled yet: read the examples)
- Fires on
- 0.73% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 87%
- python 8%
- story 2%
- chat 1%
- worksheet 1%
Makes the model say
tests):test``` togetherose "!"`. =
Makes it not say
havingOnce1print needed amDon ways
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
agent 4.0ifference` to `minus` in `math_utils.py` and
-
agent 3.98<|assistant_start|>Renamed `halve` to `halve_v2` in `utils.py` and
-
agent 3.93Fixed `max_item` in `lists.py`: it
-
agent 3.66sub_one` to `new_sub_one` in `utils.py`.<|assistant_end|>
-
agent 3.61passed⏎<|tool_end|><|assistant_start|>Fixed `decrement` in `number_utils.py`: it
-
agent 3.59last_letter` to `my_last_letter` in `script.py`, `
-
agent 3.51raise_to` to `power` in `test_number_utils.py` and
-
agent 3.51<|assistant_end|><|tool_start|>all tests passed⏎<|tool_end|><|assistant_start|>Fixed `shout` in `utils.py`: it
-
agent 3.5` to `add_one_v2` in `number_utils.py`, `
-
agent 3.44last_letter` to `my_last_letter` in `greetings.py`, `
-
agent 3.44`is_even_v2` in `test_math_utils.py`, `
-
agent 3.42passed⏎<|tool_end|><|assistant_start|>Fixed `max_of` in `number_utils.py`: it