Bug: Incorrect CORRECTNESS_PROMPT — missing format placeholders (user_question, report, answer)

[nariaki3551]

Hi team! Thank you for the great project.

I found that CORRECTNESS_PROMPT does not use user_question, report, or answer in tests/prompts.py:

open_deep_research/tests/prompts.py

Lines 95 to 104 in b419df8

	CORRECTNESS_PROMPT = """You are evaluating the correctness of a research report that was generated by a research agent.
	You will be provided with the question, the report, and the answer from an independent authority.
	Score the report from 1-5 on how well it mirrors the answer from the authority.
	We expect the report to contain more information that is not in the answer, that's perfectly okay.
	They likely won't be perfectly the same, but they should have the same themes and ideas to get a high score.
	Use your best judgement when comparing the answer to the report!
	"""

However, the eval_correctness function in tests/evaluators.py formats the prompt with these fields:

open_deep_research/tests/evaluators.py

Line 112 in b419df8

user_input_content = CORRECTNESS_PROMPT.format(user_question=query, report=final_report, answer=answer, today=get_today_str())

As a result, format() has no effect and the evaluator receives a prompt that lacks the actual question, the generated report, and the authoritative answer.

Could you please confirm whether this is a bug?