无需训练，用图文描述+AI评判实现可解释的农作物病虫害诊断。
Agri-CPJ: A Training-Free Explainable Framework for Agricultural Pest Diagnosis Using Caption-Prompt-Judge and LLM-as-a-Judge
- 先生成结构化植物特征描述，多维度筛选优化后再作诊断。
- 相比无描述基线，识别准确率提升22.7个百分点，问答得分高19.5点。
- 结果可追溯，农户能查出错误判断来自哪条观察描述，适合田间应用。
从田间照片进行作物病害诊断面临两大难题：模型在基准测试中表现良好却常误报物种名，且即使预测正确，其推理过程对从业者不可见。本文提出Agri-CPJ（Caption-Prompt-Judge）——一种无需训练的少样本框架。首先利用大视觉语言模型生成结构化形态描述，并通过多维质量筛选迭代优化；随后从互补视角生成两个候选回答，由LLM裁判依据领域标准选择更优者。消融实验证明，跳过描述优化会显著降低下游性能。在CDDMBench上，搭配GPT-5-mini生成的描述，GPT-5-Nano实现疾病分类准确率提升22.7个百分点，问答得分提高19.5点；在AgMMU-MCQs上，未修改直接适配，GPT-5-Nano达77.84%，Qwen-VL-Chat达64.54%，均优于多数同规模开源模型。结构化描述与裁判理由共同构成可读审计轨迹，使用者可定位错误判断的具体观测依据。代码与数据已公开于https://github.com/CPJ-Agricultural/CPJ-Agricultural-Diagnosis。
原文摘要 · Abstract (English)
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
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。