Human Expert Evaluation Invitation

Dear Expert:

We are currently conducting an evaluative experiment in the field of neuroscience and cordially invite you to participate. This study aims to systematically assess model-generated responses across five key dimensions: Accuracy, Reasoning, Hallucination, Coverage, and Terminology.

During the experiment, you will be asked to evaluate a set of randomly assigned records. Each record consists of a Query, a Summary (the model’s response), and the Evidence_data. Please assign an integer score from 1 to 10 for each dimension based on the following criteria:

• 10: Flawless; exceptional performance with no omissions.
• 8-9: Strong performance, with only minor imperfections or slight omissions.
• 6-7: Generally correct, but with notable weaknesses or minor errors.
• 4-5: Insufficient logical or factual reliability; contains significant errors or hallucinations.
• 1-3: Fundamentally incorrect or logically incoherent; unacceptable.

Evaluation Dimensions and Focus Areas:

• Accuracy: Assess factual correctness and consistency with neuroscience facts or common sense.
• Reasoning: Assess logical coherence and whether conclusions follow from evidence.
• Hallucination: Check whether the response contains fabricated or unsupported claims.
• Coverage: Check whether key evidence and data types are comprehensively covered.
• Terminology: Check correctness and appropriateness of domain-specific terminology.

Please adhere strictly to these definitions to ensure objective evaluation. Your scores will serve as a gold standard to verify model reliability and guide future improvements. We sincerely appreciate your participation and professional contribution.