Feature #943
Fires on โ.โ (not labeled yet: read the examples)
- Fires on
- 0.51% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 96%
- story 2%
- python 1%
- chat 0%
- worksheet 0%
Makes the model say
<|bos|>. `".โd nodY
Makes it not say
your saw herre the my times could
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
agent 2.42factorial(SIZE))<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path=hello.py<|arg|>oldString=
-
agent 2.419, 4))<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path=main.py<|arg|>oldString=
-
agent 2.41, 2))<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path=main.py<|arg|>oldString=
-
agent 2.38letter(TEXT))<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path=script.py<|arg|>oldString=
-
agent 2.34_even(SIZE))<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path=script.py<|arg|>oldString=
-
agent 2.33+ 1<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path=strings.py<|arg|>oldString=
-
agent 2.29laf(7))<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path=main.py<|arg|>oldString=
-
agent 2.28py (3 replacements)<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path=main.py<|arg|>oldString=
-
agent 2.280, 3]))<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path=main.py<|arg|>oldString=
-
agent 2.27_positive(LIMIT))<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path=main.py<|arg|>oldString=
-
agent 2.24py (2 replacements)<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path=calc.py<|arg|>oldString=
-
agent 2.24, 11))<|tool_end|><|assistant_start|><|tool_call_start|>edit<|arg|>path=main.py<|arg|>oldString=