Feature #2302
Fires on “ (” (not labeled yet: read the examples)
- Fires on
- 0.71% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 83%
- chat 8%
- story 6%
- python 1%
- worksheet 0%
Makes the model say
oldString checks //Tomonfigy nowcontent
Makes it not say
5 could2 closed words4 asked1
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
agent 2.57newString=TEXT = "apple"<|tool_call_end|><|assistant_end|><|tool_start|>Edited main.py (1 replacement)<|tool_end|>
-
agent 2.55newString=TEXT = "robot"<|tool_call_end|><|assistant_end|><|tool_start|>Edited main.py (1 replacement)<|tool_end|>
-
agent 2.53in OpenCode, a coding agent harness.<|system_end|><|user_start|>Run the tests and fix any bug.<|user_end|>
-
agent 2.53in OpenCode, a coding agent harness.<|system_end|><|user_start|>Run the tests and fix any bug.<|user_end|>
-
agent 2.52in OpenCode, a coding agent harness.<|system_end|><|user_start|>Run the tests and fix any bug.<|user_end|>
-
agent 2.5in OpenCode, a coding agent harness.<|system_end|><|user_start|>Make run.py print "It works
-
agent 2.49in OpenCode, a coding agent harness.<|system_end|><|user_start|>Run the tests and fix any bug.<|user_end|>
-
agent 2.48newString=WORD = "moon"<|tool_call_end|><|assistant_end|><|tool_start|>Edited demo.py (1 replacement)<|tool_end|>
-
agent 2.48in OpenCode, a coding agent harness.<|system_end|><|user_start|>Please main.py crashes, can
-
agent 2.48<|arg|>newString=LIMIT = 4<|tool_call_end|><|assistant_end|><|tool_start|>Edited main.py (1 replacement)<|tool_end|>
-
agent 2.47, a coding agent harness.<|system_end|><|user_start|>"Please run the tests and fix any bug."<|user_end|>
-
agent 2.46newString=WORD = "river"<|tool_call_end|><|assistant_end|><|tool_start|>Edited run.py (1 replacement)<|tool_end|>