Feature #1490
Fires on “ +” (not labeled yet: read the examples)
- Fires on
- 0.91% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 55%
- python 37%
- story 4%
- worksheet 2%
- chat 2%
Makes the model say
surprise two b longer, smell familydef
Makes it not say
nderprint len happenang<|system_start|>⏎ climb
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
agent 4.02a⏎⏎⏎def average(a, b):⏎ return (a + b) /
-
agent 3.69b⏎⏎⏎def mean(a, b):⏎ return (a + b) /
-
python 3.65`return a + b / 2` with `return (a + b) /
-
agent 3.632⏎⏎⏎def average(a, b):⏎ return (a + b) /
-
agent 3.61()⏎⏎⏎def mean(a, b):⏎ return (a + b) /
-
agent 3.6)⏎⏎⏎def mean(a, b):⏎ return (a + b) /
-
agent 3.590⏎⏎⏎def average(a, b):⏎ return (a + b) /
-
python 3.55(a + b) / 3` with `return (a + b) /
-
agent 3.52py (1 replacement)<|tool_end|><|assistant_start|>Added `min_of(a, b)` to `
-
agent 3.52.py` defines 2 functions: `max_of(a, b)` and `
-
agent 3.51py (1 replacement)<|tool_end|><|assistant_start|>Added `sum_two(a, b)` to `
-
agent 3.51return a + b / 2` instead of `return (a + b) /