Feature #2612
Fires on “ matches” (not labeled yet: read the examples)
- Fires on
- 0.79% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 95%
- story 3%
- chat 1%
- python 0%
- worksheet 0%
Makes the model say
T away winl nothing who thingscript
Makes it not say
+x python =ed exampleassert equals
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
agent 3.44quiet<|tool_call_end|><|assistant_end|><|tool_start|>Found 1 matches⏎utils.py
-
agent 3.39<|assistant_end|><|tool_start|>Found 1 matches⏎tools.py
-
agent 3.37def tail<|tool_call_end|><|assistant_end|><|tool_start|>Found 1 matches⏎listlib.py
-
agent 3.3grep<|arg|>pattern=def cube<|tool_call_end|><|assistant_end|><|tool_start|>Found 1 matches⏎math_utils
-
agent 3.28<|arg|>pattern=def absolute<|tool_call_end|><|assistant_end|><|tool_start|>Found 1 matches⏎number_utils
-
agent 3.26p<|arg|>pattern=def is_positive<|tool_call_end|><|assistant_end|><|tool_start|>Found 1 matches⏎utils.py
-
agent 3.26<|tool_call_end|><|assistant_end|><|tool_start|>Found 1 matches⏎items.py
-
agent 3.26<|tool_call_end|><|assistant_end|><|tool_start|>Found 1 matches⏎app.py
-
agent 3.26<|tool_call_end|><|assistant_end|><|tool_start|>Found 1 matches⏎stats.py
-
agent 3.26<|tool_call_end|><|assistant_end|><|tool_start|>Found 1 matches⏎app.py
-
agent 3.26grep<|arg|>pattern=def initial<|tool_call_end|><|assistant_end|><|tool_start|>Found 1 matches⏎textlib.py
-
agent 3.23p<|arg|>pattern=def max_item<|tool_call_end|><|assistant_end|><|tool_start|>Found 1 matches⏎lists.py