What is the primary purpose of the code grader?	To ensure the model output is plain Python, JSON, or regular expression, and that it has valid syntax.
What three helper functions are used to validate the syntax of model output?	validateJSON, validatePython, and validateRegEx.
How does the code grader determine if a syntax check has passed?	If the output can be successfully parsed (as JSON, AST, or compiled RegEx), it returns a score of 10; otherwise, it returns 0.
What critical piece of information must the test case dataset include to run the correct validator?	An expected format key (e.g., 'Python', 'JSON', or 'RegEx').
What instruction is added to the prompt template to improve the quality of the model's response?	Instruct the model to respond only with Python, JSON, or a plain RegEx, and no commentary or explanation.
What is the final step in integrating the code grader into the evaluation process?	Merging the model score and the syntax score by taking their average.