---

CURRENT TIME: {{CURRENT_TIME}}

Please act as a text quality evaluation expert and rate the provided query and related compact evidence objects according to the following requirements:

1. Evaluation Objectives

• A query that needs to be addressed
• A list of compact evidence objects. Each object may include source_id, title, url/source, key_passages, and publish_time. The full article body is intentionally not provided by default.

2. Rating Dimensions (10-point scale, 0 = lowest, 10 = highest)

2.1 Relevance

Measures the direct connection between the content segment and the specific topic, core concepts, and sub-questions of the query. Relevance is determined by whether the segment explicitly focuses on the query's unique subject (not just general or related fields).

• High score (8-10): Explicitly addresses the query's specific topic and core concepts; directly relates to all or part of the query's sub-questions.
• Medium score (4-7): Mentions general concepts related to the query but does not focus on the specific topic or sub-questions.
• Low score (0-3): Discusses unrelated topics with no connection to the query's specific subject, even if it shares vague keywords.

2.2 Answerability

Evaluates how effectively the segment provides direct, specific information to answer the query's sub-questions or resolve its intent. A segment's relevance does not guarantee answerability.

• High score (8-10): Provides concrete details, examples, or solutions that directly answer part or all of the query's sub-questions.
• Medium score (4-7): Offers background context related to the query but does not directly address the sub-questions or lacks specific information.
• Low score (0-3): Fails to provide any information that helps answer the query's sub-questions, even if it is topically relevant.

2.3 Authority

Assesses the authority (qualifications, professional background, field recognition) of the information source and the reliability (accuracy, objectivity, transparency of origin) of its output content. Sources with completely unidentifiable and untraceable identities (e.g., anonymous entities without any verifiable affiliation/background) should be excluded from authority and reliability evaluation.

• High score (8-10): The source has clear and authoritative qualifications; its content clearly indicates information sources and is consistent with widely verified facts; there is no obvious bias, and it has a long-term record of credible information release.
• Medium score (4-7): The source has basic professional qualifications but limited field influence; its content partially indicates information sources,
  and there is no major factual error but may contain slight subjective tendencies; its credibility record is generally good with no serious trust-breaking incidents.
• Low score (0-3): The source has no verifiable professional qualifications or relevant background; its content has no clear information source, contains obvious factual errors or extreme biases; or it has a record of fabricating information, spreading rumors, or other serious trust-breaking behaviors; or the source's authority and reliability cannot be judged by any available information, it receives ≤3.

2.4 Data Density

Measures the concentration and substantive use of empirical or quantitative evidence within the report segment. This dimension evaluates whether claims and analyses are underpinned by a meaningful volume of data—such as numerical values, systematic observations, recorded measurements, or structured datasets. A high score indicates that data is not only abundant but also strategically integrated to drive insight, support reasoning, and substantiate conclusions.

High score (8–10): Demonstrates high data density through a substantial, specific, and well-organized body of quantitative evidence—such as large-scale datasets, multiple complementary metrics, or longitudinal/time-series observations—that is central to the analysis and directly validates key claims or findings. Medium score (4–7): Contains some empirical data, but its scope, depth, variety, or integration is limited; often relies on qualitative assertions or broad interpretations that lack robust quantitative backing. Low score (0–3): Exhibits minimal or no meaningful data; primarily consists of subjective opinions, vague summaries, or non-quantitative narratives with little to no measurable or verifiable evidence.

3. Output Format Requirements

Return a JSON array where each element is a dictionary containing:

• "document_index": the index of the compact evidence document from the input documents list.
• "doc_time": According to the content of the document, extract the writen time of the document content in the format of year and month(e.g. 2023 Jun/2024 8月).
• "brief_reason": A concise reason for the scores, no more than 120 Chinese characters or 80 English words.
• "scores": A nested dictionary containing:
  • "relevance": Relevance score (10-point scale)
  • "answerability": Answerability score (10-point scale)
  • "authority": Authority score (10-point scale)
  • "data_density": Data Density score (10-point scale)

Example output format (must be pure json without any Markdown formatting):

[ { "document_index": 0, "doc_time": "2023 Jun", "brief_reason": "The evidence directly answers the query with high-density quantitative details.", "scores": { "relevance": 9.0, "answerability": 8.5, "authority": 9.0, "data_density": 9.0 } }, { "document_index": 1, "doc_time": "2024 8月", "brief_reason": "The evidence is related but only partially answers the query.", "scores": { "relevance": 7.0, "answerability": 6.5, "authority": 8.0, "data_density": 7.0 } } ]

4. Important Notes

• Strictly follow the above format; do not add any additional explanations or text.
• Ensure each compact evidence document has a corresponding rating, maintaining the same order as the input documents list.
• Evaluate consistency after comprehensively analyzing all segments, avoiding isolated assessments.

Now for the query: {{query}} please rate all the following documents: