Question about the Polyglot experiment

[Wenyi-AI-Wang]

Hi,

The get_current_edits function of coding_agent_polyglot.py is returning a message with the Claude format. However, it is stated in the paper that Polyglot experiments use O3-mini. I wonder if this function should check what language model it is using and return in the corresponding format? Or is it expected that the self-improving process can figure this out and fix it?

Also, I think line 116 of coding_agent_polyglot.py should be

self.code_model = CLAUDE_MODEL if not self_improve else OPENAI_MODEL

instead of

self.code_model = CLAUDE_MODEL if not self_improve else CLAUDE_MODEL

Thanks