[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-1607":3,"related-1607":79},{"id":4,"title":5,"url":6,"summary":7,"summary_zh":8,"source_name":9,"source_url":6,"published_at":10,"category":11,"cover_url":12,"hotness":13,"is_selected":14,"embedding_ready":15,"score":16,"sources":17,"tags":19,"primary_topic":23,"secondary_topics":29,"search_phrases":30,"slug":33,"view_count":34,"doi":35,"paper":36,"created_at":54,"content":12,"score_detail":55,"classification":62},1607,"Artificial intelligence and agricultural production efficiency","https:\u002F\u002Fdoi.org\u002F10.3389\u002Ffsufs.2026.1886625","As a core driving force of the new round of technological revolution, artificial intelligence (AI) is profoundly reshaping agricultural production modes and the pattern of factor allocation. However, the effect of AI on agricultural production efficiency is not a linear process. The mechanisms through which this effect operates, and how these mechanisms evolve as technology penetration deepens, remain theoretical and practical questions that require urgent answers. Using panel data for 31 Chinese provincial-level regions from 2005 to 2023, this paper applies a multi-period difference-in-differences (DID) method to systematically examine the impact of AI on agricultural production efficiency and the dynamics of mechanism switching. Three findings emerge. First, AI development has a significant positive effect on agricultural production efficiency. Second, this effect is transmitted through two mechanisms: land use efficiency and smart technology adoption. Third, the two mechanisms exhibit a sequential switch in dominance, following a pattern in which smart technology adoption dominates first and land use efficiency improvement dominates later. This study extends static mechanism testing to a dynamic analysis of mechanism switching. It provides a theoretical basis and a quantitative reference for designing differentiated agricultural digitalization policies.","作为新一轮科技革命的核心驱动力，人工智能（AI）正深刻重塑农业生产方式与要素配置格局。然而，AI对农业生产效率的影响并非线性过程，其作用机制如何传导，以及这些机制如何随着技术渗透的深化而演变，仍是亟待回答的理论与实践问题。本文利用2005—2023年中国31个省级行政区的面板数据，采用多期双重差分（DID）方法，系统考察了AI对农业生产效率的影响及其机制切换的动态特征。研究发现：第一，AI发展对农业生产效率具有显著的正向促进作用；第二，这一效应通过土地利用效率与智能技术采纳两条机制传导；第三，两条机制在主导性上呈现依次切换的规律，表现为先由智能技术采纳主导，后由土地利用效率提升主导。本研究将静态机制检验拓展至机制切换的动态分析，为设计差异化的农业数字化政策提供了理论依据与量化参考。","Frontiers in Sustainable Food Systems","2026-09-03T00:00:00Z","论文",null,10,false,true,78,[18],{"name":9,"url":6},[20,21,22],"智慧农业","农业人工智能","生产效率",{"code":24,"slug":25,"name":26,"path":27,"parent_code":28},"I2","agricultural-ai","农业AI与大模型","\u002Ftopic\u002Fdigital-agriculture\u002Fagricultural-ai","I",[],[31,32],"农业人工智能 智慧农业 生产效率","农业人工智能 智慧农业","农业人工智能智慧农业生产效率-1607",1,"10.3389\u002Ffsufs.2026.1886625",{"doi":35,"openalex_id":37,"authors":38,"venue":9,"cited_by_count":45,"oa_url":46,"card":47,"direction":51,"ingested_from":53},"W7207663052",[39,41,43],{"name":40,"orcid":12},"Yi Zhou",{"name":42,"orcid":12},"Kuntong Tang",{"name":44,"orcid":12},"Baocong Han",0,"https:\u002F\u002Fwww.frontiersin.org\u002Fjournals\u002Fsustainable-food-systems\u002Farticles\u002F10.3389\u002Ffsufs.2026.1886625\u002Fpdf",{"tldr":48,"method":49,"finding":50,"direction":51,"opportunity":52},"利用中国省级面板数据，实证检验AI对农业生产效率的影响及机制转换。","多期DID方法，分析2005-2023年中国31省面板数据。","AI显著提升农业生产效率，机制从智能技术采用转向土地利用效率。","农业人工智能与决策模型","可探索不同农业发展阶段下AI影响机制的动态转换条件，以及区域异质性政策设计。","openalex","2026-09-04T23:30:07.910501Z",{"impact":56,"substance":57,"depth":58,"authority":59,"freshness":60,"relevant":34,"comment":61},20,22,18,13,5,"基于省级面板数据与多期DID方法，揭示AI对农业效率的双机制动态切换，为政策设计提供量化参考。",{"status":63,"locked":14,"version":64,"method":65,"evidence":66},"classified","2026-10-22.2","rules",{"candidates":67,"agricultural_evidence":15},[68,74],{"code":24,"score":69,"supported":15,"title":70,"abstract":72,"tags":73},12,[71],"artificial intelligence",[71],[21],{"code":75,"score":34,"supported":14,"title":76,"abstract":77,"tags":78},"I1",[],[],[20],{"total":80,"page":34,"page_size":80,"items":81},6,[82,127,184,226,264,298],{"id":83,"title":84,"url":85,"summary":86,"summary_zh":87,"source_name":88,"source_url":85,"published_at":89,"category":11,"cover_url":12,"hotness":13,"is_selected":14,"embedding_ready":15,"score":90,"sources":91,"tags":93,"primary_topic":97,"secondary_topics":98,"search_phrases":104,"slug":107,"view_count":45,"doi":108,"paper":109,"created_at":126},5613,"ExplainableCropAI: Interpretable Deep Learning for Agricultural Disease Diagnosis and Decision Support","https:\u002F\u002Fdoi.org\u002F10.1007\u002Fs44196-026-01654-9","Deep learning models classify crop diseases accurately, but their black-box nature limits practical use: farmers and extension officers need to understand a diagnosis before acting on it. We present ExplainableCropAI, a decision-support system for paddy disease diagnosis that attaches three complementary explanation channels to a ConvNeXt V2-Tiny classifier: Grad-CAM++ visual attribution, Monte Carlo Dropout uncertainty with temperature-scaled calibration, and natural-language explanations generated by Phi-3-mini under a grounding constraint that restricts the model to typed fields drawn from the prediction, the attribution map and a curated agronomic knowledge base. None of the components is new; the contribution lies in their integration into a single pipeline and in measuring the properties that integration is meant to deliver. On a leakage-controlled, held-out test split of 1,540 images from the Paddy Doctor dataset (10,407 images, nine diseases and healthy), the classifier reaches 97.5% accuracy (95% CI 96.7–98.3) and macro-F1 of 96.9%. Six attribution methods are compared under identical settings with model-based faithfulness metrics; the calibrated confidence separates correct from incorrect predictions with an AUROC of 0.93, and referring the 10% least-confident cases to an expert removes 74% of all errors. Under eight image corruptions at five severities the model retains 89.7% mean accuracy against 72–78% for six baselines trained identically, and its referral rate rises with the degree of degradation. Zero-shot evaluation on two independently collected rice-disease datasets from Bangladesh and India shows that accuracy roughly halves under cross-dataset shift and that confidence no longer identifies errors, while fifty local images per class recover accuracy above 90%. The code, split files, trained weights and evaluation scripts are released.","深度学习模型能够准确分类作物病害，但其黑箱性质限制了实际应用：农民和推广人员需要在理解诊断结果后才能采取行动。我们提出ExplainableCropAI，一个用于水稻病害诊断的决策支持系统，它在ConvNeXt V2-Tiny分类器上附加了三条互补的解释通道：Grad-CAM++视觉归因、经温度缩放校准的蒙特卡洛Dropout不确定性，以及由Phi-3-mini在接地约束下生成的自然语言解释——该约束将模型限制为从预测结果、归因图和精选农学知识库中提取的类型化字段。各组件均非新创；贡献在于将其集成到单一流水线中，并测量该集成旨在提供的特性。在来自Paddy Doctor数据集（10,407张图像，九种病害及健康）的1,540张图像上，经泄漏控制的留出测试划分显示，分类器达到97.5%的准确率（95% CI 96.7–98.3）和96.9%的宏F1值。在相同设置下，使用基于模型的忠实性指标比较了六种归因方法；校准后的置信度以0.93的AUROC区分正确与错误预测，将置信度最低的10%案例转交专家可消除全部错误的74%。在五种严重程度的八种图像损坏下，模型保持89.7%的平均准确率，而相同训练的六个基线为72–78%，且其转交率随退化程度上升。在来自孟加拉国和印度的两个独立采集的水稻病害数据集上进行零样本评估显示，跨数据集偏移下准确率大约减半，置信度不再能识别错误，而每类五十张本地图像可将准确率恢复至90%以上。代码、划分文件、训练权重和评估脚本均已发布。","International Journal of Computational Intelligence Systems","2026-10-22T00:00:00Z",84,[92],{"name":88,"url":85},[20,21,94,95,96],"水稻","可解释AI","作物病害诊断",{"code":24,"slug":25,"name":26,"path":27,"parent_code":28},[99],{"code":100,"slug":101,"name":94,"path":102,"parent_code":103},"A1","rice","\u002Ftopic\u002Ffield-crops\u002Frice","A",[105,106],"ExplainableCropAI 水稻病害","Paddy Doctor 数据集","ExplainableCropAI水稻病害-5613","10.1007\u002Fs44196-026-01654-9",{"doi":108,"openalex_id":110,"authors":111,"venue":88,"cited_by_count":45,"oa_url":85,"card":121,"direction":51,"ingested_from":53},"W7166554515",[112,114,115,118,120],{"name":113,"orcid":12},"Santosh Kumar",{"name":113,"orcid":12},{"name":116,"orcid":117},"Durgesh Nandan","https:\u002F\u002Forcid.org\u002F0000-0002-9762-3559",{"name":119,"orcid":12},"Sumit Kumar",{"name":119,"orcid":12},{"tldr":122,"method":123,"finding":124,"direction":51,"opportunity":125},"构建可解释水稻病害诊断系统，融合归因、不确定性与自然语言解释。","ConvNeXt V2-Tiny分类器，Grad-CAM++、MC Dropou","测试准确率97.5%，低置信转诊消除74%错误，跨数据集精度减半。","跨数据集置信失效，少量本地样本微调可恢复精度，值得研究田间自适应与主动学习。","2026-10-22T23:31:14.582261Z",{"id":128,"title":129,"url":130,"summary":131,"summary_zh":132,"source_name":133,"source_url":130,"published_at":134,"category":11,"cover_url":12,"hotness":13,"is_selected":14,"embedding_ready":15,"score":135,"sources":136,"tags":138,"primary_topic":142,"secondary_topics":143,"search_phrases":149,"slug":152,"view_count":45,"doi":153,"paper":154,"created_at":183},5603,"AI-Enabled Smart and Precision Agriculture: Smart Sensors, IoT, Deep Learning, Computer Vision, and Decision Support: A Comprehensive Survey","https:\u002F\u002Fdoi.org\u002F10.20944\u002Fpreprints202610.2112.v1","Artificial intelligence (AI) is increasingly integrated with smart sensors, Internet of Things (IoT) networks, machine learning, deep learning, and computer vision to support site-specific agricultural management. However, strong performance on computational benchmarks does not necessarily translate into reliable decisions, effective field interventions, or measurable improvements in agricultural productivity and resource efficiency. This comprehensive narrative survey critically synthesizes advances in agricultural sensing, IoT architectures, data processing, multimodal sensor fusion, predictive modeling, computer vision, decision optimization, and automated actuation. Particular emphasis is placed on the relationships between measurement reliability, model generalization, decision feasibility, intervention accuracy, and verified agricultural outcomes. The reviewed literature demonstrates promising capabilities in crop monitoring, yield prediction, disease and weed recognition, precision irrigation, and selected automated operations. Nevertheless, heterogeneous evaluation protocols, limited independent field validation, sensor calibration challenges, distribution shifts, interoperability barriers, and deployment costs constrain operational reliability and generalizability. To address these interconnected challenges, the survey proposes a literature-informed reference architecture linking agricultural observations, data integration, AI inference, constrained decision-making, physical interventions, and outcome feedback, supported by uncertainty assessment, human oversight, and cross-layer assurance. The proposed architecture is conceptual and has not been empirically validated. Future research priorities include reproducible field evaluations, representative agricultural datasets, uncertainty-aware and explainable AI, resource-efficient computing, interoperable infrastructures, and economically inclusive deployment. The review concludes that the practical value of AI-enabled precision agriculture should be evaluated through reliable interventions, sustainable resource utilization, and demonstrated agricultural outcomes rather than predictive performance alone.","人工智能（AI）正日益与智能传感器、物联网（IoT）网络、机器学习、深度学习及计算机视觉相融合，以支撑因地制宜的精准农业管理。然而，在计算基准上的优异表现并不必然转化为可靠的决策、有效的田间干预，或农业生产率与资源效率方面可衡量的提升。本综述性研究系统梳理并批判性综合了农业传感、物联网架构、数据处理、多模态传感器融合、预测建模、计算机视觉、决策优化及自动执行等方面的进展，重点探讨测量可靠性、模型泛化能力、决策可行性、干预精度与经核实的农业成效之间的关系。所综述文献表明，上述技术在作物监测、产量预测、病害与杂草识别、精准灌溉及部分自动化作业中展现出良好前景。然而，评估方案异质、独立田间验证有限、传感器校准难题、分布偏移、互操作性障碍及部署成本等因素制约了运行可靠性与泛化能力。为应对这些相互关联的挑战，本综述提出一种基于文献的参考架构，将农业观测、数据集成、AI推理、受约束决策、物理干预与成效反馈相衔接，并以不确定性评估、人工监督与跨层保障为支撑。该架构为概念性框架，尚未经实证验证。未来研究重点包括可复现的田间评估、具有代表性的农业数据集、不确定性感知与可解释AI、资源高效计算、可互操作基础设施及经济包容性部署。本综述认为，AI赋能精准农业的实际价值应通过可靠的干预措施、可持续的资源利用及经证实的农业成效来评估，而非仅凭预测性能。","Preprints.org","2026-10-21T00:00:00Z",70,[137],{"name":133,"url":130},[20,21,139,140,141],"计算机视觉","物联网","精准农业",{"code":24,"slug":25,"name":26,"path":27,"parent_code":28},[144],{"code":145,"slug":146,"name":147,"path":148,"parent_code":28},"I5","agricultural-iot","物联网传感与边缘计算","\u002Ftopic\u002Fdigital-agriculture\u002Fagricultural-iot",[150,151],"农业人工智能 计算机视觉 智慧农业 精准农业","农业人工智能 计算机视觉","农业人工智能计算机视觉智慧农业精准农业-5603","10.20944\u002Fpreprints202610.2112.v1",{"doi":153,"openalex_id":155,"authors":156,"venue":133,"cited_by_count":45,"oa_url":130,"card":177,"direction":181,"ingested_from":53},"W7223175252",[157,160,163,166,168,170,173,175],{"name":158,"orcid":159},"IBRAHIM ADABARA","https:\u002F\u002Forcid.org\u002F0000-0001-8949-7540",{"name":161,"orcid":162},"Bashir Olaniyi Sadiq","https:\u002F\u002Forcid.org\u002F0000-0001-7868-4076",{"name":164,"orcid":165},"Aliyu Nuhu Shuaibu","https:\u002F\u002Forcid.org\u002F0000-0002-5171-675X",{"name":167,"orcid":12},"Yale Ibrahim Danjuma",{"name":169,"orcid":12},"Ainamukama Anold",{"name":171,"orcid":172},"Mutebi Joe","https:\u002F\u002Forcid.org\u002F0000-0002-2757-3875",{"name":174,"orcid":12},"Elhadi Abdalla A. Arbab",{"name":176,"orcid":12},"Brian Mugisha",{"tldr":178,"method":179,"finding":180,"direction":181,"opportunity":182},"综述AI赋能精准农业的传感、物联网、深度学习与决策支持，并提出参考架构。","叙述性综述，整合传感器、IoT、多模态融合、计算机视觉与决策优化文献。","基准性能不等于田间可靠决策，评估协议异质、独立验证不足制约落地。","智慧农业 \u002F 农业物联网","可研究不确定性感知与可解释AI、可复现田间评估及跨层保障架构的实证验证。","2026-10-22T23:30:19.310160Z",{"id":185,"title":186,"url":187,"summary":188,"summary_zh":189,"source_name":190,"source_url":187,"published_at":134,"category":11,"cover_url":12,"hotness":13,"is_selected":14,"embedding_ready":15,"score":191,"sources":192,"tags":194,"primary_topic":198,"secondary_topics":204,"search_phrases":206,"slug":209,"view_count":45,"doi":210,"paper":211,"created_at":225},5592,"Evaluation of Concept-Rank Collapse Audits for Computer Vision Agriculture Disease Classifiers Under Low-Label Regimes: An Interpretability Capacity Bound for Smallholder Fungicide Selectivity","https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.23545193","Concept-based interpretability methods for plant-disease classifiers are typically evaluated by whether their concept activations correlate with human-annotated labels, but this criterion is silent on a prior question: whether the classifier's internal representation retains enough concept-rank structure for any audit to be informative. the present analysis formalises that prior question as an interpretability capacity bound. A classifier is modelled as a concept-rank operator whose singular values govern how much concept-level signal survives the projection from pixel space to class logits, and a collapse audit is defined as the ratio of retained concept-rank energy to total concept energy. Under a low-label regime, where concept annotations are scarce and the audit must be estimated from few labelled examples, the bound predicts that audit reliability degrades as a function of the concept-rank spectrum and the label budget. The analysis is a theoretical and methodological contribution: it derives the bound, states the conditions under which a collapse audit is identifiable, and characterises the operating characteristics of the audit as a decision procedure under a declared variance structure (between-cluster SD = 6.0, residual SD = 4.5, intraclass correlation = 0.64). Power was estimated across 2000 design cells for cluster counts of 8, 16, 24 and 32 and effect sizes of 3, 5 and 8 percentage points; 52 clusters were required for 5 percentage points at 80% power. The bound reframes concept audits as capacity tests rather than fidelity tests, with consequences for how smallholder fungicide-selectivity decisions should be supported by interpretable disease classifiers.","面向植物病害分类器的基于概念的可解释性方法，通常以其概念激活是否与人工标注标签相关作为评价标准，但该标准对一个更先前的问题保持沉默：分类器的内部表征是否保留了足够的概念秩结构，从而使任何审计具有信息量。本分析将该先前问题形式化为一个可解释性容量上界。分类器被建模为概念秩算子，其奇异值支配着从像素空间到类别logits的投影中概念级信号得以保留的程度，并将坍缩审计定义为保留的概念秩能量与总概念能量之比。在低标签情形下，概念标注稀缺且审计必须由少量标注样本估计，该上界预测审计可靠性会随概念秩谱和标签预算的函数而退化。本分析是一项理论和方法论贡献：推导出该上界，陈述坍缩审计可识别的条件，并在声明的方差结构（簇间SD = 6.0，残差SD = 4.5，组内相关系数 = 0.64）下刻画审计作为决策程序的操作特征。在2000个设计单元中，针对簇数为8、16、24和32以及效应量为3、5和8个百分点的情形估计了功效；在80%功效下达到5个百分点需要52个簇。该上界将概念审计重新框定为容量检验而非保真度检验，并对可解释病害分类器应如何支持小农杀菌剂选择性决策具有启示意义。","Zenodo (CERN European Organization for Nuclear Research)",69,[193],{"name":190,"url":187},[20,21,195,196,197],"小农户","植物病害识别","可解释性",{"code":199,"slug":200,"name":201,"path":202,"parent_code":203},"G2","farm-services","社会化服务与规模经营","\u002Ftopic\u002Frural-services\u002Ffarm-services","G",[205],{"code":24,"slug":25,"name":26,"path":27,"parent_code":28},[207,208],"农业人工智能 植物病害识别 可解释性 智慧农业","农业人工智能 植物病害识别","农业人工智能植物病害识别可解释性智慧农业-5592","10.5281\u002Fzenodo.23545193",{"doi":210,"openalex_id":212,"authors":213,"venue":190,"cited_by_count":45,"oa_url":187,"card":220,"direction":51,"ingested_from":53},"W7223137323",[214,216,218],{"name":215,"orcid":12},"Akshay Ramgoolam",{"name":217,"orcid":12},"Priya Bundhoo Appadoo",{"name":219,"orcid":12},"Jean-Noël LEspérance",{"tldr":221,"method":222,"finding":223,"direction":51,"opportunity":224},"提出可解释性容量上界，评估低标签下概念秩崩溃审计的可靠性。","将分类器建模为概念秩算子，推导容量上界并做2000格功效模拟。","审计可靠性随概念秩谱与标签预算下降；5个百分点效应需52个聚类达80%功效。","可将容量上界用于小农户杀菌剂选择性决策的可解释模型审计与标签预算设计。","2026-10-22T11:21:10.425290Z",{"id":227,"title":228,"url":229,"summary":230,"summary_zh":231,"source_name":190,"source_url":229,"published_at":134,"category":11,"cover_url":12,"hotness":13,"is_selected":14,"embedding_ready":15,"score":191,"sources":232,"tags":234,"primary_topic":238,"secondary_topics":244,"search_phrases":246,"slug":249,"view_count":45,"doi":250,"paper":251,"created_at":263},5591,"Hierarchical Conformal Risk Control for Artificial Intelligence Agriculture Irrigation Scheduling Under Unmodeled Infiltration: A Computer Safety-Guarantee Evaluation","https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.23543665","Irrigation scheduling algorithms that rely on soil-moisture forecasts are vulnerable to unmodeled infiltration, the process by which water bypasses the root zone through preferential pathways that resistance-based soil models cannot represent. The framework is evaluated through a replication-based design study that characterises the operating characteristics of the before-after-control-impact estimator used to audit scheduler performance, under a declared variance structure with between-cluster standard deviation 6.0, residual standard deviation 4.5 and intraclass correlation 0.64. Power was estimated across 2000 design cells spanning cluster counts of 8, 16, 24 and 32 and effect sizes of 3, 5 and 8 percentage points. The analysis indicates that 54 clusters are required to detect a 5-percentage-point effect at 80% power, and that the minimum detectable effect declines steeply as cluster count rises from 8 to 32. The contribution is a formal safety guarantee for irrigation decision support that remains valid when the infiltration model is wrong, together with an audit design that makes the guarantee empirically checkable.","依赖土壤湿度预报的灌溉调度算法容易受到未建模入渗的影响，即水分通过优先路径绕过根区的过程，而基于阻力的土壤模型无法表征这一过程。该框架通过一项基于重复的设计研究进行评估，该研究刻画了用于审计调度器性能的前后对照影响估计量的运行特性，其所依据的声明方差结构为：组间标准差6.0、残差标准差4.5、组内相关系数0.64。功效估计覆盖2000个设计单元，涵盖聚类数8、16、24和32以及效应量3、5和8个百分点。分析表明，在80%功效下检测5个百分点效应需要54个聚类，且最小可检测效应随聚类数从8增至32而急剧下降。本文的贡献在于为灌溉决策支持提供了一种正式的安全保证，该保证在入渗模型错误时仍然有效，同时提出了一种使该保证可经实证检验的审计设计。",[233],{"name":190,"url":229},[20,21,235,236,237],"灌溉决策","安全保证","试验设计",{"code":239,"slug":240,"name":241,"path":242,"parent_code":243},"H3","irrigation","水资源与灌溉","\u002Ftopic\u002Fenvironment\u002Firrigation","H",[245],{"code":24,"slug":25,"name":26,"path":27,"parent_code":28},[247,248],"灌溉调度 安全保证","农业人工智能 安全保证 智慧农业 灌溉决策","灌溉调度安全保证-5591","10.5281\u002Fzenodo.23543665",{"doi":250,"openalex_id":252,"authors":253,"venue":190,"cited_by_count":45,"oa_url":229,"card":258,"direction":51,"ingested_from":53},"W7223131205",[254,256],{"name":255,"orcid":12},"Nassima Benkhelifa",{"name":257,"orcid":12},"Yacine Aït Ahmed",{"tldr":259,"method":260,"finding":261,"direction":51,"opportunity":262},"提出分层保形风险控制，为未建模入渗下的灌溉调度提供安全保证。","分层保形风险控制与基于重复设计的BACI审计，2000个设计单元功效分析。","检测5个百分点效应需54个集群，最小可检测效应随集群数增加陡降。","可将保形风险控制扩展到其他未建模扰动下的农业决策安全保证与审计设计。","2026-10-22T11:21:10.283561Z",{"id":265,"title":266,"url":267,"summary":268,"summary_zh":269,"source_name":190,"source_url":267,"published_at":134,"category":11,"cover_url":12,"hotness":13,"is_selected":14,"embedding_ready":15,"score":135,"sources":270,"tags":272,"primary_topic":275,"secondary_topics":276,"search_phrases":278,"slug":281,"view_count":45,"doi":282,"paper":283,"created_at":297},5590,"Evaluation of Concept-Probe Temporal-Stability Audits for Computer Vision Agriculture Disease Classifiers Under Seasonal Phenology Drift: An Interpretability Decay Bound for Smallholder Fungicide Timing","https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.23545234","Concept-probe audits are increasingly used to certify that a crop-disease classifier's internal representations track the pathological concepts a smallholder would act on, but a probe certified at one point in the season may silently decouple from those concepts as phenology shifts the visual statistics of the field. the present analysis treats that decoupling as a formal object. It defines a temporal-stability audit as a sequence of concept-probe accuracies measured across defined phenological windows, and it derives an interpretability decay bound: the rate at which probe accuracy can fall before the fungicide-timing decision it is meant to support becomes unreliable. The bound is expressed as a function of the probe's initial margin, the drift in concept prevalence between windows, and the classifier's sensitivity to the drifted features, and it yields a certified audit interval rather than a single-point certificate. The paper then characterises the operating characteristics of the audit design under a declared variance structure, using a replication-based evaluation of the before-after-control-impact estimator with between-cluster SD = 6.0, residual SD = 4.5 and intraclass correlation = 0.64. Power was estimated across 2000 design cells over cluster counts of 8, 16, 24 and 32 and effect sizes of 3, 5 and 8 percentage points; 54 clusters were required for 5 percentage points at 80% power. The contribution is a decision-theoretic account of when a concept-probe certificate expires, and a design rule for scheduling re-audits across the season.","概念探针审计（concept-probe audit）日益被用于验证作物病害分类器的内部表征是否追踪小农户会据以采取行动的病理学概念，但在生长季某一时点获得认证的探针，可能随着物候变化改变田间的视觉统计特征而悄然与这些概念脱钩。本分析将这种脱钩视为一个形式化对象。它将时间稳定性审计（temporal-stability audit）定义为在既定物候窗口内测量的一系列概念探针准确率，并推导出一个可解释性衰减界限（interpretability decay bound）：即探针准确率在其所支持的杀菌剂施用时机决策变得不可靠之前所能下降的速率。该界限表示为探针初始裕度、窗口间概念流行率漂移以及分类器对漂移特征敏感度的函数，并给出一个经认证的审计区间而非单点认证。随后，本文在声明的方差结构下刻画该审计设计的运行特征，采用基于重复的前后对照影响估计量评估，其中组间标准差 = 6.0、残差标准差 = 4.5、组内相关系数 = 0.64。功效在 2000 个设计单元上估计，聚类数取 8、16、24 和 32，效应量取 3、5 和 8 个百分点；在 80% 功效下达到 5 个百分点需要 54 个聚类。本文的贡献在于从决策论角度阐明概念探针认证何时失效，并给出在整个生长季内安排重新审计的设计规则。",[271],{"name":190,"url":267},[20,21,195,273,274],"作物病害识别","模型审计",{"code":199,"slug":200,"name":201,"path":202,"parent_code":203},[277],{"code":24,"slug":25,"name":26,"path":27,"parent_code":28},[279,280],"概念探针 时序稳定性 作物病害分类器","杀菌剂 施药时机 可解释性衰减","概念探针时序稳定性作物病害分类器-5590","10.5281\u002Fzenodo.23545234",{"doi":282,"openalex_id":284,"authors":285,"venue":190,"cited_by_count":45,"oa_url":267,"card":292,"direction":51,"ingested_from":53},"W7223088030",[286,288,290],{"name":287,"orcid":12},"Abdi-Basid Idriss Farah",{"name":289,"orcid":12},"Amina Abdillahi Guedi",{"name":291,"orcid":12},"Moussa Omar Youssouf",{"tldr":293,"method":294,"finding":295,"direction":51,"opportunity":296},"提出概念探针时间稳定性审计与可解释性衰减界，判断何时需重新审计。","定义跨物候窗口探针精度序列，推导衰减界，用BACI复制评估统计功效。","5个百分点效应需54个聚类达80%功效，审计证书会随物候漂移过期。","可将衰减界与再审计调度规则嵌入小农户杀菌剂时机决策支持系统，并验证跨作物泛化。","2026-10-22T11:21:08.978348Z",{"id":299,"title":300,"url":301,"summary":302,"summary_zh":303,"source_name":190,"source_url":301,"published_at":134,"category":11,"cover_url":12,"hotness":13,"is_selected":14,"embedding_ready":15,"score":304,"sources":305,"tags":307,"primary_topic":310,"secondary_topics":311,"search_phrases":312,"slug":315,"view_count":45,"doi":316,"paper":317,"created_at":327},5589,"Conformal-Risk Evaluation for Artificial Intelligence Agriculture Spray-Nozzle Fault Detection Under Label-Delay Feedback: A Computer Coverage Audit","https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.23544107","Spray-nozzle fault detection in precision agriculture is a classification problem in which the label that confirms a fault arrives after the spray pass, so the detector must act on unlabelled evidence and its error rate must be certified before the label exists. Conformal prediction offers distribution-free coverage guarantees, but the guarantee is stated for exchangeable calibration and test data, an assumption that label-delayed feedback violates because the calibration set is always older than the deployment set. the present analysis formalises label-delay feedback as a distribution shift in the conformity score and derives a coverage bound for split-conformal fault detectors under that shift. The bound separates the nominal miscoverage rate into an exchangeable component and a delay-induced component that scales with the total variation distance between the calibration and deployment score distributions. A replication-based evaluation of the ['Intercept', 'exposure', 'group', 'severity'] estimator was conducted: 600 data sets of N = 500 were drawn from a declared data-generating process (intercept 12.0, exposure 0.42, group -3.1, severity 1.35, error SD 4.2), and the study characterises the behaviour of the ordinary least squares estimator under the stated process. The estimator recovered the declared coefficients with estimated bias 0.01122 for the intercept and -0.0002 for exposure, root mean squared error 0.9217 and 0.01614 respectively, and empirical coverage of the nominal 95% intervals of 1.0 for both parameters across 600 replications. The coverage audit shows that the delay-induced term, not the exchangeable term, dominates the risk of a conformal spray-nozzle detector, and that the audit procedure can certify a detector before field labels return.","在精准农业中，喷雾喷嘴故障检测是一个分类问题，其中确认故障的标签在喷雾作业结束后才到达，因此检测器必须基于未标记的证据采取行动，并且其错误率必须在标签存在之前得到认证。保形预测（conformal prediction）提供了无分布覆盖保证，但该保证是针对可交换的校准数据和测试数据而言的，而标签延迟反馈违反了这一假设，因为校准集总是比部署集更旧。本分析将标签延迟反馈形式化为一致性分数中的分布偏移，并推导了在该偏移下分裂保形故障检测器的覆盖界。该界将名义错误覆盖 rate 分解为可交换分量和延迟诱导分量，后者随校准与部署分数分布之间的全变差距离而缩放。对 ['Intercept', 'exposure', 'group', 'severity'] 估计器进行了基于重复的评估：从声明的数据生成过程（截距 12.0，暴露 0.42，分组 -3.1，严重程度 1.35，误差标准差 4.2）中抽取了 600 个 N = 500 的数据集，研究刻画了普通最小二乘估计器在所述过程下的行为。该估计器恢复了声明的系数，截距的估计偏差为 0.01122，暴露为 -0.0002，均方根误差分别为 0.9217 和 0.01614，两个参数在 600 次重复中名义 95% 区间的经验覆盖率为 1.0。覆盖审计表明，延迟诱导项而非可交换项主导了保形喷雾喷嘴检测器的风险，并且该审计程序可以在田间标签返回之前对检测器进行认证。",62,[306],{"name":190,"url":301},[20,21,308,309],"精准施药","病虫害监测",{"code":24,"slug":25,"name":26,"path":27,"parent_code":28},[],[313,314],"农业人工智能 病虫害监测 智慧农业 精准施药","农业人工智能 病虫害监测","农业人工智能病虫害监测智慧农业精准施药-5589","10.5281\u002Fzenodo.23544107",{"doi":316,"openalex_id":318,"authors":319,"venue":190,"cited_by_count":45,"oa_url":301,"card":322,"direction":51,"ingested_from":53},"W7223116087",[320],{"name":321,"orcid":12},"Achieng' Atim Okello",{"tldr":323,"method":324,"finding":325,"direction":51,"opportunity":326},"将标签延迟反馈建模为共形分数分布偏移，推导覆盖率界并审计喷雾喷嘴故障检测器。","分裂共形预测、总变差距离、600次重复模拟（N=500）评估OLS估计。","延迟诱导项而非可交换项主导检测器风险，审计可在田间标签返回前认证检测器。","可将该覆盖率审计框架扩展到其他标签延迟的农业传感器故障检测，并研究延迟分布估计误差对认证的影响。","2026-10-22T11:21:02.845718Z"]