Feature #3342
Fires on “py” (not labeled yet: read the examples)
- Fires on
- 0.88% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 98%
- story 1%
- chat 0%
- python 0%
- worksheet 0%
Makes the model say
<|arg|><|tool_call_end|><|system_end|>Sure): %pattern sentence
Makes it not say
s⏎ count They moral Soon Every say
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
agent 2.47<|arg|>replaceAll=true<|tool_call_end|><|assistant_end|><|tool_start|>Edited test_mathlib.py (2 replacements)
-
agent 2.47rue<|tool_call_end|><|assistant_end|><|tool_start|>Edited textlib.py (1 replacement)
-
agent 2.45<|assistant_start|><|tool_call_start|>glob<|arg|>pattern=**/*.py<|tool_call_end|><|assistant_end|><|tool_start|>main.py<|tool_end|><|assistant_start|>This project
-
agent 2.34(1, n + 1):<|tool_call_end|><|assistant_end|><|tool_start|>Edited utils.py (1 replacement)
-
agent 2.31n % 2 == 0]<|tool_call_end|><|assistant_end|><|tool_start|>Edited stats.py (1 replacement)
-
agent 2.3<|arg|>replaceAll=true<|tool_call_end|><|assistant_end|><|tool_start|>Edited test_tools.py (2 replacements)
-
agent 2.3return -x<|tool_call_end|><|assistant_end|><|tool_start|>Edited mathlib.py (1 replacement)
-
agent 2.28=greet<|arg|>replaceAll=true<|tool_call_end|><|assistant_end|><|tool_start|>Edited main.py (2 replacements)
-
agent 2.27v2<|arg|>replaceAll=true<|tool_call_end|><|assistant_end|><|tool_start|>Edited run.py (2 replacements)
-
agent 2.22<|assistant_start|><|tool_call_start|>glob<|arg|>pattern=**/*<|tool_call_end|><|assistant_end|><|tool_start|>words.py⏎test_lib
-
agent 2.22s):⏎ return s.upper()<|tool_call_end|><|assistant_end|><|tool_start|>Edited textlib.py (1 replacement)
-
agent 2.22<|arg|>replaceAll=true<|tool_call_end|><|assistant_end|><|tool_start|>Edited test_calc.py (2 replacements)