数据与计算发展前沿 ›› 2026, Vol. 8 ›› Issue (2): 171-183.

CSTR: 32002.14.jfdc.CN10-1649/TP.2026.02.013

doi: 10.11871/jfdc.issn.2096-742X.2026.02.013

• 技术与应用 • 上一篇    下一篇

基于多模型协作与动态加权裁决的农业科学元数据智能审核系统框架

任有强1,2,3(),赵慧1,2,李威1,2,3,袁欢1,樊景超1,2,3,*(),张建华1,2,3,4,周国民2,5,6   

  1. 1 中国农业科学院农业信息研究所/农业农村部农业大数据重点实验室北京 100081
    2 国家农业科学数据中心北京 100081
    3 三亚中国农业科学院国家南繁研究院海南 三亚 572024
    4 海南省种业实验室海南 三亚 572024
    5 农业农村部南京农业机械化研究所江苏 南京 210014
    6 中国农业科学院西部农业研究中心新疆 昌吉 831100
  • 收稿日期:2025-07-17 出版日期:2026-04-20 发布日期:2026-04-23
  • 通讯作者: *樊景超(E-mail: fanjingchao@caas.cn
  • 作者简介:任有强,中国农业科学院农业信息研究所,硕士研究生,研究方向为农业信息技术。
    本文中主要工作为提出研究思路,设计实验方案,构建系统,撰写论文。
    REN Youqiang is a master’s student at the Agricultural Information Institute of the Chinese Academy of Agricultural Sciences (CAAS). His research focuses on agricultural information technology.
    In this paper, he is responsible for proposing the research ideas, designing the experimental scheme, constructing the system, and writing the manuscript.
    E-mail:17864179721@163.com|樊景超,中国农业科学院农业信息研究所,博士,副研究员,研究方向为农业大数据。
    本文中主要工作为设计研究方案,实验结果的解读与分析。
    FAN Jingchao, Ph.D., is an associate researcher at the Agricultural Information Institute of the Chinese Academy of Agricultural Sciences (CAAS). His research focuses on agricultural big data.
    In this study, he is responsible for designing the research scheme and analyzing and interpreting the experimental results.
    E-mail: fanjingchao@caas.cn
  • 基金资助:
    国家重点研发计划(2022YFF0711800);国家重点研发计划(2022YFD1600300);海南省自然科学基金(325MS155);三亚崖州湾科技城科技专项资助(SCKJ-JYRC-2023-45);三亚中国农业科学院国家南繁研究院南繁专项(YBXM2430);三亚中国农业科学院国家南繁研究院南繁专项(YBXM2508);三亚中国农业科学院国家南繁研究院南繁专项(YBXM2509);中央级公益性科研院所基本科研业务费专项(JBYW-AII-2025-05);中央级公益性科研院所基本科研业务费专项(Y2025YC90);国家农业科学数据中心项目(NASDC2025XM11);三亚崖州湾科技城管理局海南省种业实验室2025年产业科技创新“揭榜挂帅”联合项目(B25H1JC14)

A Framework for Intelligent Auditing of Agricultural Science Metadata Based on Multi-Model Collaboration and Dynamic Weighted Adjudication

REN Youqiang1,2,3(),ZHAO Hui1,2,LI Wei1,2,3,YUAN Huan1,FAN Jingchao1,2,3,*(),ZHANG Jianhua1,2,3,4,ZHOU Guomin2,5,6   

  1. 1 Institute of Agricultural Information, Chinese Academy of Agricultural Sciences / Key Laboratory of Agricultural Big Data, Ministry of Agriculture and Rural Affairs, Beijing 100081, China
    2 National Agricultural Science Data Center, Beijing 100081, China
    3 Nanfan Research Institute (Sanya), Chinese Academy of Agricultural Sciences, Sanya, Hainan 572024, China
    4 Hainan Provincial Seed Industry Laboratory, Sanya, Hainan 572024, China
    5 Nanjing Institute of Agricultural Mechanization, Ministry of Agriculture and Rural Affairs, Nanjing, Jiangsu 210014, China
    6 Western Research Institute, Chinese Academy of Agricultural Sciences, Changji, Xinjiang 831100, China
  • Received:2025-07-17 Online:2026-04-20 Published:2026-04-23

摘要:

【目的】 为解决农业元数据人工审核效率低、标准不一的技术问题,提出一种面向质量治理的智能审核框架。【方法】 该框架采用两阶段分层策略:第一阶段利用多种异构大语言模型(LLMs)并行执行初审,并结合“算法三角测量”从语义、词汇、结构三维度评估输出一致性;第二阶段引入动态加权系统总分歧度(DWSD)算法,量化模型间冲突并动态触发高性能裁决模型,实现高难度样本的精确审核。【结果】 在真实农业元数据集上,该框架在F1分数、精确率和召回率上较最佳基线模型分别提升15.24%、14.17%和15.00%,显著优化了审核准确性与覆盖率。【局限】研究在单一中文农业数据集上验证,其跨语言、跨领域的泛化能力有待检验;所应用的偏见缓解策略在更大规模数据下的有效性也需持续评估。【结论】 本研究提出的智能框架可有效提升农业元数据审核的智能化、精确化与可解释性,并为多语言、多领域的数据质量治理任务提供可推广的技术路径。

关键词: 农业元数据审核, 多模型协作, LLM-as-a-Judge, 动态加权, DWSD

Abstract:

[Objective] This study proposes a quality governance-oriented intelligent auditing framework to address inefficiency, inconsistent standards in agricultural science metadata review. [Methods] The framework adopts a two-stage hierarchical strategy: (1) Multiple heterogeneous Large Language Models (LLMs) perform parallel preliminary audits, evaluated through an “algorithmic triangulation” method across semantic, lexical, and structural dimensions; (2) A Dynamic Weighted System Disagreement (DWSD) algorithm quantifies inter-model conflicts and dynamically triggers high-performance adjudicator models for precise handling of challenging cases. [Results] On a real-world agricultural metadata dataset, the proposed framework improved F1-score, precision, and recall by 15.24%, 14.17%, and 15.00%, respectively, over the best baseline, significantly enhancing audit accuracy and coverage. [Limitations] The model was validated on a single Chinese agricultural dataset; its generalizability across different languages and domains requires further testing. Additionally, the effectiveness of the applied bias mitigation strategies at a larger scale needs ongoing assessment. [Conclusions] The intelligent framework proposed in this study can effectively enhance the intelligence, accuracy, and interpretability of agricultural metadata auditing, and provide a generalizable technical path for data quality governance tasks across multiple languages and domains.

Key words: agricultural metadata audit, multi-model collaboration, LLM-as-a-Judge, Dynamic Weighting, DWSD