arXiv 2604.237012026-04-26cs.CLcs.AIcs.CV

Agri-CPJ：一种无需训练的可解释框架，用于利用Caption-Prompt-Judge和LLM-as-a-Judge进行农业害虫诊断

Agri-CPJ: A Training-Free Explainable Framework for Agricultural Pest Diagnosis Using Caption-Prompt-Judge and LLM-as-a-Judge

Business School, Shandong University of Technology（山东理工大学商学院）；Faculty of Data Science, City University of Macau（澳门城市大学数据科学学院）；Guangdong Institute of Intelligent Science and Technology（广东智能科学与技术研究院）；Macau Millennium College（澳门 millennium 学院）；Department of Computer and Information Science, University of Macau（澳门大学计算机与信息科学系）

Wentao Zhang, Qi Zhang, Mingkun Xu, Mu You, Henghua Shen, Zhongzhi He, Keyan Jin, Derek F. Wong, Tao Fang

备注：: This work is an expanded version of our prior paper published in the IEEE ICASSP 2026 conference arXiv:2512.24947, from 4 to 20+ pages, presenting a well-structured and principled framework, extensive experiments, and deeper insights. Tao Fang is the corresponding author

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出Agri-CPJ框架，通过生成结构化形态描述并利用LLM进行判断，解决农业病害诊断中模型易产生错误物种名称和推理不可用的问题，实验显示其在病害分类和问答评分上有显著提升。

摘要：

农田病害诊断面临两个问题：在基准测试中表现良好的模型常产生物种名称的幻觉，且即使预测正确，推理过程也对从业者不可用。本文描述了Agri-CPJ（Caption-Prompt-Judge），一种无需训练的少样本框架，其中大视觉-语言模型首先生成结构化的形态描述，通过多维质量门控逐步优化，再生成两个互补的响应，由LLM判断者根据领域特定标准选择更强的响应。描述优化是影响最大的组件：消融实验确认跳过它会一致降低下游准确性。在CDDMBench上，将GPT-5-Nano与GPT-5-mini生成的描述结合，病害分类准确率提高22.7个百分点，问答评分提高19.5分。在未修改的情况下评估AgMMU-MCQs，GPT-5-Nano达到77.84%，Qwen-VL-Chat达到64.54%，尽管格式从开放式转换为多项选择，仍处于大多数同类开源模型之上。结构化描述和判断理由共同构成可读的审计轨迹：对诊断有异议的从业者可以识别出具体的描述观察错误。代码和数据已公开：https://github.com/CPJ-Agricultural/CPJ-Agricultural-Diagnosis

英文摘要：

Crop disease diagnosis from field photographs faces two recurring problems: models that score well on benchmarks frequently hallucinate species names, and when predictions are correct, the reasoning behind them is typically inaccessible to the practitioner. This paper describes Agri-CPJ (Caption-Prompt-Judge), a training-free few-shot framework in which a large vision-language model first generates a structured morphological caption, iteratively refined through multi-dimensional quality gating, before any diagnostic question is answered. Two candidate responses are then generated from complementary viewpoints, and an LLM judge selects the stronger one based on domain-specific criteria. Caption refinement is the component with the largest individual impact: ablations confirm that skipping it consistently degrades downstream accuracy across both models tested. On CDDMBench, pairing GPT-5-Nano with GPT-5-mini-generated captions yields \textbf{+22.7} pp in disease classification and \textbf{+19.5} points in QA score over no-caption baselines. Evaluated without modification on AgMMU-MCQs, GPT-5-Nano reached 77.84\% and Qwen-VL-Chat reached 64.54\%, placing them at or above most open-source models of comparable scale despite the format shift from open-ended to multiple-choice. The structured caption and judge rationale together constitute a readable audit trail: a practitioner who disagrees with a diagnosis can identify the specific caption observation that was incorrect. Code and data are publicly available https://github.com/CPJ-Agricultural/CPJ-Agricultural-Diagnosis

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。