[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-56":3,"related-56":52},{"id":4,"title":5,"url":6,"summary":7,"summary_zh":8,"content":9,"source_name":10,"source_url":8,"published_at":11,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":15,"score_detail":16,"sources":24,"tags":26,"primary_topic":8,"secondary_topics":33,"classification":34,"search_phrases":47,"slug":50,"view_count":20,"doi":8,"paper":8,"created_at":51},56,"Agri-SAGE：基于模拟的多智能体大语言模型用于上下文感知农业咨询生成","https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.00454","Agri-SAGE是一个闭环框架，旨在通过将检索增强的多智能体LLM推理与基于APSIM的生物物理模拟相结合，生成并验证农艺建议，从而解决静态农艺指南\"对季节内变化和动态不确定性视而不见\"以及LLM驱动咨询系统\"生成农艺上可信但生理上不可信的建议\"两个核心矛盾。研究评估了三种推理方法——规划与求解、思维树和反思法，进行了为期10年的回顾性分析。结果显示三种推理方法均优于静态基线，其中思维树实现最高产量，反思法以较低计算成本达到相当效果。",null,"[Skip to main content](https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.00454#content)\n\n![Image 1](https:\u002F\u002Farxiv.org\u002Fstatic\u002Fbase\u002F1.0.1\u002Fimages\u002Ficons\u002Fsmileybones-small.svg)arXiv is now an independent nonprofit![Learn more](https:\u002F\u002Finfo.arxiv.org\u002Fabout)×\n\n[](https:\u002F\u002Farxiv.org\u002FIgnoreMe)[![Image 2: archive](https:\u002F\u002Farxiv.org\u002Fstatic\u002Fbase\u002F1.0.1\u002Fimages\u002Farxiv-logo-primary-light.svg)](https:\u002F\u002Farxiv.org\u002F)[Search](https:\u002F\u002Farxiv.org\u002Fsearch)[Submit](https:\u002F\u002Farxiv.org\u002Fuser\u002Fcreate)[Donate](https:\u002F\u002Finfo.arxiv.org\u002Fabout\u002Fdonate.html)[Log in](https:\u002F\u002Farxiv.org\u002Flogin)\n\nSearch arXiv \n\n Press Enter to search · [Advanced search](https:\u002F\u002Farxiv.org\u002Fsearch\u002Fadvanced)\n\n# Computer Science > Artificial Intelligence\n\n**arXiv:2607.00454** (cs) \n\n [Submitted on 1 Jul 2026]\n\n# Title:Agri-SAGE: Simulation-Grounded Multi-Agent LLM for Context-Aware Agricultural Advisory Generation\n\nAuthors:[Vedant Balasubramaniam](https:\u002F\u002Farxiv.org\u002Fsearch\u002Fcs?searchtype=author&query=Balasubramaniam,+V), [Geetha Charan](https:\u002F\u002Farxiv.org\u002Fsearch\u002Fcs?searchtype=author&query=Charan,+G), [Manojkumar Patil](https:\u002F\u002Farxiv.org\u002Fsearch\u002Fcs?searchtype=author&query=Patil,+M), [Rohit P Suresh](https:\u002F\u002Farxiv.org\u002Fsearch\u002Fcs?searchtype=author&query=Suresh,+R+P), [V Priyanka](https:\u002F\u002Farxiv.org\u002Fsearch\u002Fcs?searchtype=author&query=Priyanka,+V), [Kodur Sai Vinay Sathvik](https:\u002F\u002Farxiv.org\u002Fsearch\u002Fcs?searchtype=author&query=Sathvik,+K+S+V), [Y. Narahari](https:\u002F\u002Farxiv.org\u002Fsearch\u002Fcs?searchtype=author&query=Narahari,+Y)\n\nView a PDF of the paper titled Agri-SAGE: Simulation-Grounded Multi-Agent LLM for Context-Aware Agricultural Advisory Generation, by Vedant Balasubramaniam and Geetha Charan and Manojkumar Patil and Rohit P Suresh and V Priyanka and Kodur Sai Vinay Sathvik and Y. Narahari\n\n[View PDF](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2607.00454)[HTML (experimental)](https:\u002F\u002Farxiv.org\u002Fhtml\u002F2607.00454v1)\n> Abstract:Agricultural advisory systems face a fundamental tension: static agronomic guidelines offer consistent, evidence-based recommendations, yet remain blind to in-season variability and dynamic uncertainties. Recent advisory systems powered by LLMs are liable for a different risk of generating recommendations that are agronomically credible but physiologically unconvincing. Agri-SAGE is a closed-loop framework designed to resolve the above two limitations by integrating retrieval-grounded multi-agent LLM reasoning with APSIM-based biophysical simulation, to generate and validate agronomic advisories. To assess this framework, we evaluate three reasoning approaches, namely Plan-and-Solve, Tree of Thoughts, and Reflexion, over a 10-year retrospective analysis. All three significantly outperform static PoP (Package-of-Practice) baselines, with Tree of Thoughts achieving impressive peak yields. At the same time, Reflexion achieves comparable agronomic outcomes at substantially lower computational cost by leveraging cross-seasonal episodic memory.\n\nSubjects:Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)\nCite as:[arXiv:2607.00454](https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.00454) [cs.AI]\n(or [arXiv:2607.00454v1](https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.00454v1) [cs.AI] for this version)\n[https:\u002F\u002Fdoi.org\u002F10.48550\u002FarXiv.2607.00454](https:\u002F\u002Fdoi.org\u002F10.48550\u002FarXiv.2607.00454)\n\nFocus to learn more\n\n arXiv-issued DOI via DataCite\n\n## Submission history\n\n From: Geetha Charan [[view email](https:\u002F\u002Farxiv.org\u002Fshow-email\u002Fe08e6969\u002F2607.00454)] \n\n**[v1]** Wed, 1 Jul 2026 05:18:20 UTC (583 KB)\n\n[](https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.00454)Full-text links:\n## Access Paper:\n\n View a PDF of the paper titled Agri-SAGE: Simulation-Grounded Multi-Agent LLM for Context-Aware Agricultural Advisory Generation, by Vedant Balasubramaniam and Geetha Charan and Manojkumar Patil and Rohit P Suresh and V Priyanka and Kodur Sai Vinay Sathvik and Y. Narahari\n\n*   [View PDF](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2607.00454)\n*   [HTML (experimental)](https:\u002F\u002Farxiv.org\u002Fhtml\u002F2607.00454v1)\n*   [TeX Source](https:\u002F\u002Farxiv.org\u002Fsrc\u002F2607.00454)\n\n[![Image 3: license icon](https:\u002F\u002Farxiv.org\u002Ficons\u002Flicenses\u002Fby-4.0.png)view license](http:\u002F\u002Fcreativecommons.org\u002Flicenses\u002Fby\u002F4.0\u002F \"Rights to this article\")\n\n### Current browse context:\n\ncs.AI\n\n[\u003Cprev](https:\u002F\u002Farxiv.org\u002Fprevnext?id=2607.00454&function=prev&context=cs.AI \"previous in cs.AI (accesskey p)\") | [next>](https:\u002F\u002Farxiv.org\u002Fprevnext?id=2607.00454&function=next&context=cs.AI \"next in cs.AI (accesskey n)\")\n\n[new](https:\u002F\u002Farxiv.org\u002Flist\u002Fcs.AI\u002Fnew) | [recent](https:\u002F\u002Farxiv.org\u002Flist\u002Fcs.AI\u002Frecent) | [2026-07](https:\u002F\u002Farxiv.org\u002Flist\u002Fcs.AI\u002F2026-07)\n\n Change to browse by: \n\n[cs](https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.00454?context=cs)\n\n[cs.MA](https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.00454?context=cs.MA)\n\n### References & Citations\n\n*   [NASA ADS](https:\u002F\u002Fui.adsabs.harvard.edu\u002Fabs\u002FarXiv:2607.00454)\n*   [Google Scholar](https:\u002F\u002Fscholar.google.com\u002Fscholar_lookup?arxiv_id=2607.00454)\n*   [Semantic Scholar](https:\u002F\u002Fapi.semanticscholar.org\u002FarXiv:2607.00454)\n\nexport BibTeX citation Loading...\n\n## BibTeX formatted citation\n\n×\n\nData provided by: [](https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.00454)\n\n### Bookmark\n\n[![Image 4: BibSonomy](https:\u002F\u002Farxiv.org\u002Fstatic\u002Fbrowse\u002F0.3.4\u002Fimages\u002Ficons\u002Fsocial\u002Fbibsonomy.png)](http:\u002F\u002Fwww.bibsonomy.org\u002FBibtexHandler?requTask=upload&url=https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.00454&description=Agri-SAGE:%20Simulation-Grounded%20Multi-Agent%20LLM%20for%20Context-Aware%20Agricultural%20Advisory%20Generation \"Bookmark on BibSonomy\")[![Image 5: Reddit](https:\u002F\u002Farxiv.org\u002Fstatic\u002Fbrowse\u002F0.3.4\u002Fimages\u002Ficons\u002Fsocial\u002Freddit.png)](https:\u002F\u002Freddit.com\u002Fsubmit?url=https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.00454&title=Agri-SAGE:%20Simulation-Grounded%20Multi-Agent%20LLM%20for%20Context-Aware%20Agricultural%20Advisory%20Generation \"Bookmark on Reddit\")\n\nBibliographic Tools \n\n# Bibliographic and Citation Tools\n\n- [x] Bibliographic Explorer Toggle \n\nBibliographic Explorer _([What is the Explorer?](https:\u002F\u002Finfo.arxiv.org\u002Flabs\u002Fshowcase.html#arxiv-bibliographic-explorer))_\n\n- [x] Connected Papers Toggle \n\nConnected Papers _([What is Connected Papers?](https:\u002F\u002Fwww.connectedpapers.com\u002Fabout))_\n\n- [x] Litmaps Toggle \n\nLitmaps _([What is Litmaps?](https:\u002F\u002Fwww.litmaps.co\u002F))_\n\n- [x] scite.ai Toggle \n\nscite Smart Citations _([What are Smart Citations?](https:\u002F\u002Fwww.scite.ai\u002F))_\n\nCode, Data, Media \n\n# Code, Data and Media Associated with this Article\n\n- [x] alphaXiv Toggle \n\nalphaXiv _([What is alphaXiv?](https:\u002F\u002Falphaxiv.org\u002F))_\n\n- [x] Links to Code Toggle \n\nCatalyzeX Code Finder for Papers _([What is CatalyzeX?](https:\u002F\u002Fwww.catalyzex.com\u002F))_\n\n- [x] DagsHub Toggle \n\nDagsHub _([What is DagsHub?](https:\u002F\u002Fdagshub.com\u002F))_\n\n- [x] GotitPub Toggle \n\nGotit.pub _([What is GotitPub?](http:\u002F\u002Fgotit.pub\u002Ffaq))_\n\n- [x] Huggingface Toggle \n\nHugging Face _([What is Huggingface?](https:\u002F\u002Fhuggingface.co\u002Fhuggingface))_\n\n- [x] ScienceCast Toggle \n\nScienceCast _([What is ScienceCast?](https:\u002F\u002Fsciencecast.org\u002Fwelcome))_\n\nDemos \n\n# Demos\n\n- [x] Replicate Toggle \n\nReplicate _([What is Replicate?](https:\u002F\u002Freplicate.com\u002Fdocs\u002Farxiv\u002Fabout))_\n\n- [x] Spaces Toggle \n\nHugging Face Spaces _([What is Spaces?](https:\u002F\u002Fhuggingface.co\u002Fdocs\u002Fhub\u002Fspaces))_\n\n- [x] Spaces Toggle \n\nTXYZ.AI _([What is TXYZ.AI?](https:\u002F\u002Ftxyz.ai\u002F))_\n\nRelated Papers \n\n# Recommenders and Search Tools\n\n- [x] Link to Influence Flower \n\nInfluence Flower _([What are Influence Flowers?](https:\u002F\u002Finfluencemap.cmlab.dev\u002F))_\n\n- [x] Core recommender toggle \n\nCORE Recommender _([What is CORE?](https:\u002F\u002Fcore.ac.uk\u002Fservices\u002Frecommender))_\n\n*   [Author](https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.00454)\n*   [Venue](https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.00454)\n*   [Institution](https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.00454)\n*   [Topic](https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.00454)\n\n About arXivLabs  \n\n# arXivLabs: experimental projects with community collaborators\n\narXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.\n\nBoth individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.\n\nHave an idea for a project that will add value for arXiv's community? [**Learn more about arXivLabs**](https:\u002F\u002Finfo.arxiv.org\u002Flabs\u002Findex.html).\n\n[Which authors of this paper are endorsers?](https:\u002F\u002Farxiv.org\u002Fauth\u002Fshow-endorsers\u002F2607.00454) | [Disable MathJax](javascript:setMathjaxCookie()) ([What is MathJax?](https:\u002F\u002Finfo.arxiv.org\u002Fhelp\u002Fmathjax.html)) \n\n We gratefully acknowledge support from our **major funders**, [**member institutions**](https:\u002F\u002Finfo.arxiv.org\u002Fabout\u002Fourmembers.html), , and all contributors. \n\n[About](https:\u002F\u002Finfo.arxiv.org\u002Fabout)·[Help](https:\u002F\u002Finfo.arxiv.org\u002Fhelp)·[Contact](https:\u002F\u002Finfo.arxiv.org\u002Fhelp\u002Fcontact.html)·[Subscribe](https:\u002F\u002Finfo.arxiv.org\u002Fhelp\u002Fsubscribe)·[Copyright](https:\u002F\u002Finfo.arxiv.org\u002Fhelp\u002Flicense\u002Findex.html)·[Privacy](https:\u002F\u002Finfo.arxiv.org\u002Fhelp\u002Fpolicies\u002Fprivacy_policy.html)·[Accessibility](https:\u002F\u002Finfo.arxiv.org\u002Fhelp\u002Fweb_accessibility.html)·[Operational Status (opens in new tab)](https:\u002F\u002Fstatus.arxiv.org\u002F)\n\nMajor funding support from\n\n[![Image 6: Simons Foundation](https:\u002F\u002Farxiv.org\u002Fstatic\u002Fbase\u002F1.0.1\u002Fimages\u002Ffunders\u002Fsimons-foundation.png)](https:\u002F\u002Fwww.simonsfoundation.org\u002F)[![Image 7: Simons Foundation International](https:\u002F\u002Farxiv.org\u002Fstatic\u002Fbase\u002F1.0.1\u002Fimages\u002Ffunders\u002Fsimons-foundation-international.png)](https:\u002F\u002Fwww.sfi.org.bm\u002F)[![Image 8: Schmidt Sciences](https:\u002F\u002Farxiv.org\u002Fstatic\u002Fbase\u002F1.0.1\u002Fimages\u002Ffunders\u002Fschmidt-sciences.png)](https:\u002F\u002Fwww.schmidtsciences.org\u002F)","arXiv:2607.00454v1 [cs.AI] · 2026-07-02","2026-07-01T20:00:00Z","论文",10,false,76,{"impact":17,"substance":18,"depth":19,"authority":20,"freshness":21,"relevant":22,"comment":23},20,22,18,14,2,1,"提出结合多智能体LLM与APSIM仿真的农业咨询框架，方法新颖，但论文发布于近一个月前，时效性低。",[25],{"name":10,"url":6},[12,27,28,29,30,31,32],"预印本","农业咨询","大语言模型","多智能体","APSIM","模拟",[],{"status":35,"locked":14,"version":36,"method":37,"evidence":38},"unclassified","2026-10-22.2","rules",{"candidates":39,"agricultural_evidence":45,"reason":46},[40],{"code":41,"score":21,"supported":14,"title":42,"abstract":43,"tags":44},"I2",[],[],[30,29],true,"缺少足够的主题证据",[48,49],"大语言模型 农业咨询 多智能体 预印本","大语言模型 农业咨询","大语言模型农业咨询多智能体预印本-56","2026-07-28T23:53:31.334067Z",{"total":53,"page":22,"page_size":53,"items":54},6,[55,102,185,268,342,412],{"id":56,"title":57,"url":58,"summary":59,"summary_zh":8,"content":60,"source_name":61,"source_url":8,"published_at":62,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":63,"score_detail":64,"sources":69,"tags":71,"primary_topic":75,"secondary_topics":81,"classification":82,"search_phrases":96,"slug":99,"view_count":100,"doi":8,"paper":8,"created_at":101},58,"Persona-as-Configuration：面向农业洪涝灾害的生成式利益相关方报告框架（arXiv:2607.17774v1）","https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.17774","论文提出Persona-as-Configuration生成式报告框架，针对农业洪涝灾害场景中需要同时向农民、地方政府、保险公司、应急管理部门等多类利益相关方传递差异化信息的痛点。框架通过将利益相关方的需求特征参数化配置，结合大语言模型生成定制化、结构化的灾害报告文本，提高沟通效率与决策支持效果。","[View PDF](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2607.17774)[HTML (experimental)](https:\u002F\u002Farxiv.org\u002Fhtml\u002F2607.17774v1)\n\n> Abstract:Cyber-physical systems built on deterministic edge inference, such as on-vehicle flood detection for agricultural fields, produce structured decision logs that must be interpreted differently by heterogeneous stakeholders. Pairing such systems with large language models (LLMs) to generate stakeholder-specific reports introduces a tension: the generative layer is non-deterministic, while the edge plane must remain replayable and auditable. We propose an architectural pattern resting on two invariants: unidirectional consumption, in which the generative layer is a strict read-only consumer of the deterministic plane and never writes back, and persona-as-configuration, in which stakeholder adaptation is a versioned prompt-template artifact rather than runtime improvisation. We instantiate the pattern as a context-aware dashboard layer over the JSON decision logs of a previously published edge-based standing-water detection system, and analyse how the integration boundary admits standard generative-reliability mitigations as configuration- or middleware-level extension points. A structured expert review rated the pattern favourably across five ISO\u002FIEC 25010-aligned quality dimensions, with strongest agreement on separation of concerns. End-user evaluation with agricultural stakeholders is planned for future work.\n\nComments:Accepted at CASA 2026 (9th Workshop on Context-Aware, Autonomous and Smart Architectures), co-located with ECSA 2026. 16 pages, 2 figures\nSubjects:Software Engineering (cs.SE); Artificial Intelligence (cs.AI); Human-Computer Interaction (cs.HC)\nCite as:[arXiv:2607.17774](https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.17774) [cs.SE]\n(or [arXiv:2607.17774v1](https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.17774v1) [cs.SE] for this version)\n[https:\u002F\u002Fdoi.org\u002F10.48550\u002FarXiv.2607.17774](https:\u002F\u002Fdoi.org\u002F10.48550\u002FarXiv.2607.17774)\n\narXiv-issued DOI via DataCite (pending registration)\n\n## Submission history\n\nFrom: Oliver Larsen [[view email](https:\u002F\u002Farxiv.org\u002Fshow-email\u002F6f78a517\u002F2607.17774)] \n\n**[v1]** Mon, 20 Jul 2026 10:06:47 UTC (1,331 KB)","arXiv:2607.17774v1 · 2026-07-20","2026-07-20T00:00:00Z",69,{"impact":65,"substance":17,"depth":19,"authority":66,"freshness":67,"relevant":22,"comment":68},15,12,4,"提出面向农业洪涝灾害的生成式利益相关方报告框架，方法新颖，但时效性较差。",[70],{"name":61,"url":58},[12,27,29,72,73,74],"农业洪涝","利益相关方","报告生成",{"code":76,"slug":77,"name":78,"path":79,"parent_code":80},"H7","climate","气候适应与防灾减灾","\u002Ftopic\u002Fenvironment\u002Fclimate","H",[],{"status":83,"locked":14,"version":36,"method":37,"evidence":84},"classified",{"candidates":85,"agricultural_evidence":45},[86,92],{"code":76,"score":87,"supported":45,"title":88,"abstract":90,"tags":91},11,[89],"洪涝",[89],[],{"code":41,"score":22,"supported":14,"title":93,"abstract":94,"tags":95},[],[],[29],[97,98],"利益相关方 大语言模型 农业洪涝 报告生成","利益相关方 大语言模型","利益相关方大语言模型农业洪涝报告生成-58",8,"2026-07-28T23:53:31.370076Z",{"id":103,"title":104,"url":105,"summary":106,"summary_zh":107,"content":8,"source_name":108,"source_url":105,"published_at":109,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":15,"score_detail":110,"sources":115,"tags":117,"primary_topic":122,"secondary_topics":128,"classification":133,"search_phrases":156,"slug":159,"view_count":160,"doi":161,"paper":162,"created_at":184},5376,"Agri-Synergy: A Hardware–Software Co-Designed Hybrid IoT System for Sensor Interface Standardization and LLM-Driven Interactive Decision Support","https:\u002F\u002Fdoi.org\u002F10.3390\u002Fagriengineering8100446","This paper presents Agri-Synergy, a hybrid Internet of Things (IoT) system featuring a hardware–software co-design to tackle sensor data heterogeneity and the barriers to effective AI-driven decision support in smart agriculture. Unlike software-based data harmonization, we propose a universal sensor printed circuit board (PCB) with RS485\u002FModbus interface for hardware unification, complemented by on-node edge processing that ensures standardized data from source. To lower the entry barrier for farmers, the system integrates the DeepSeek large language model (LLM) API, combining static agricultural knowledge with real-time sensor data, for context-aware, natural language-based decision support, functioning without local model deployment. Field deployment showed reference-verified sensor accuracy predominantly within manufacturer specifications, stable 4G connectivity during the 4-h benchmark but intermittent dropouts during the 20-day deployment, and remote-control success rates of 96% (normal) and 80% (weak signal at 300 m). Multi-node concurrent transmission resulted in average latency of 238.0 ± 27.1 ms and packet loss below 0.5%. The system achieves a 3–5× cost reduction compared to commercial platforms. These results demonstrate that addressing data heterogeneity from source through hardware unification and software edge processing, while leveraging AI to lower the barrier to system adoption, provides a more reliable foundation for data synchronization and accessible AI-driven analytics.","本文提出Agri-Synergy，一种采用软硬件协同设计的混合物联网（IoT）系统，旨在解决智慧农业中传感器数据异构性问题以及有效AI驱动决策支持的障碍。与基于软件的数据协调方式不同，我们提出了一种带有RS485\u002FModbus接口的通用传感器印刷电路板（PCB）以实现硬件统一，并辅以节点端边缘处理，确保从源头获得标准化数据。为降低农民的使用门槛，该系统集成了DeepSeek大语言模型（LLM）API，将静态农业知识与实时传感器数据相结合，提供上下文感知的、基于自然语言的决策支持，且无需本地部署模型即可运行。现场部署表明，经参考验证的传感器精度基本处于制造商规格范围内；在4小时基准测试期间4G连接稳定，但在20天部署期间出现间歇性断连；远程控制成功率分别为96%（正常信号）和80%（300 m处弱信号）。多节点并发传输的平均延迟为238.0 ± 27.1 ms，丢包率低于0.5%。与商业平台相比，该系统实现了3–5倍的成本降低。这些结果表明，通过硬件统一和软件边缘处理从源头解决数据异构性问题，同时利用AI降低系统采用门槛，可为数据同步和可及的AI驱动分析提供更可靠的基础。","AgriEngineering","2026-10-19T00:00:00Z",{"impact":111,"substance":112,"depth":19,"authority":66,"freshness":113,"relevant":22,"comment":114},16,21,9,"硬件统一接口与边缘处理结合大模型决策支持的农业物联网系统论文，方法新颖、有实地部署数据，但属细分技术进展，公共影响有限。",[116],{"name":108,"url":105},[118,119,120,29,121],"智慧农业","农业人工智能","农业物联网","传感器标准化",{"code":123,"slug":124,"name":125,"path":126,"parent_code":127},"I5","agricultural-iot","物联网传感与边缘计算","\u002Ftopic\u002Fdigital-agriculture\u002Fagricultural-iot","I",[129],{"code":41,"slug":130,"name":131,"path":132,"parent_code":127},"agricultural-ai","农业AI与大模型","\u002Ftopic\u002Fdigital-agriculture\u002Fagricultural-ai",{"status":83,"locked":14,"version":36,"method":37,"evidence":134},{"candidates":135,"agricultural_evidence":45},[136,143,149],{"code":123,"score":19,"supported":45,"title":137,"abstract":139,"tags":142},[138],"IoT",[140,141,138],"物联网","internet of things",[120],{"code":41,"score":111,"supported":45,"title":144,"abstract":146,"tags":148},[145],"decision support",[147,145],"large language model",[119,29],{"code":150,"score":151,"supported":45,"title":152,"abstract":153,"tags":155},"I1",7,[],[118,154],"smart agriculture",[118],[157,158],"Agri-Synergy 物联网 传感器","DeepSeek 决策支持","Agri-Synergy物联网传感器-5376",0,"10.3390\u002Fagriengineering8100446",{"doi":161,"openalex_id":163,"authors":164,"venue":108,"cited_by_count":160,"oa_url":105,"card":177,"direction":181,"ingested_from":183},"W7222920424",[165,167,169,171,173,175],{"name":166,"orcid":8},"Leqi Lai",{"name":168,"orcid":8},"Langtao Duan",{"name":170,"orcid":8},"Yuhao Chen",{"name":172,"orcid":8},"Yuhang Zheng",{"name":174,"orcid":8},"Jinyu Yang",{"name":176,"orcid":8},"Mingyao Wen",{"tldr":178,"method":179,"finding":180,"direction":181,"opportunity":182},"提出软硬件协同的混合物联网系统，统一传感器接口并用大模型提供交互式决策支持。","通用RS485\u002FModbus传感器PCB、节点边缘处理、DeepSeek大模型A","传感器精度达标，多节点延迟238ms、丢包低于0.5%，成本降低3-5倍。","智慧农业 \u002F 农业物联网","可研究弱信号下通信可靠性提升与边缘-云端LLM协同的长期田间鲁棒性。","openalex","2026-10-20T23:30:08.319698Z",{"id":186,"title":187,"url":188,"summary":189,"summary_zh":190,"content":8,"source_name":191,"source_url":188,"published_at":192,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":193,"score_detail":194,"sources":196,"tags":198,"primary_topic":201,"secondary_topics":202,"classification":203,"search_phrases":250,"slug":253,"view_count":160,"doi":254,"paper":255,"created_at":267},5232,"Current applications of artificial intelligence for advancing agriculture, sustainability and climate mitigation","https:\u002F\u002Fdoi.org\u002F10.1007\u002Fs44163-026-02540-1","Artificial intelligence (AI) is reshaping agriculture by enabling data-driven decision-making across crop production systems and supporting the transition toward more sustainable, resilient, and climate-smart farming. This review critically synthesizes advances in AI applications spanning precision irrigation, nutrient and soil management, yield prediction, pest and disease monitoring, computer vision, autonomous agricultural robotics, and large language model (LLM)-based decision support systems. Beyond summarizing technological developments, the review evaluates their practical deployment, emphasizing that successful agricultural AI depends not only on predictive accuracy but also on model generalizability, explainability, economic feasibility, interoperability, and integration with existing agronomic knowledge. Emerging technologies including Vision Transformers, foundation models, Digital Twins, multimodal AI, Retrieval-Augmented Generation, and federated and continual learning—are discussed as important directions for overcoming limitations related to data scarcity, environmental variability, and scalable field implementation. The review highlights that responsible AI adoption requires transparent governance frameworks addressing data ownership, privacy, algorithmic fairness, cybersecurity, and equitable access, particularly for smallholder farming systems. Although numerous AI models report outstanding performance, widespread agricultural adoption remains constrained by cross-region validation, fragmented datasets, infrastructure requirements, and insufficient real-world deployment. Future progress will depend on multidisciplinary collaboration integrating advances in artificial intelligence, sensing technologies, robotics, crop science, and agricultural engineering to develop robust, trustworthy, and farmer-centered decision-support systems. By critically evaluating technological innovations and implementation challenges, this review provides a framework for understanding the current state of AI in agriculture while identifying research priorities that will shape the next generation of sustainable and climate-resilient agricultural systems.","人工智能（AI）正在重塑农业，通过在作物生产系统中实现数据驱动的决策，并支持向更可持续、更具韧性和气候智能型农业转型。本综述批判性地综合了AI应用的进展，涵盖精准灌溉、养分与土壤管理、产量预测、病虫害监测、计算机视觉、自主农业机器人以及基于大语言模型（LLM）的决策支持系统。除总结技术发展外，本综述还评估了其实际部署情况，强调农业AI的成功不仅取决于预测准确性，还取决于模型泛化能力、可解释性、经济可行性、互操作性以及与现有农学知识的整合。文中讨论了包括视觉Transformer、基础模型、数字孪生、多模态AI、检索增强生成以及联邦学习和持续学习在内的新兴技术，将其视为克服数据稀缺、环境变异性和田间可扩展实施等局限的重要方向。本综述强调，负责任的AI采用需要透明的治理框架，以解决数据所有权、隐私、算法公平性、网络安全和公平获取等问题，尤其是对小农耕作系统而言。尽管众多AI模型报告了卓越的性能，但农业领域的广泛采用仍受限于跨区域验证、碎片化数据集、基础设施要求以及现实世界部署不足。未来的进展将取决于多学科协作，整合人工智能、传感技术、机器人技术、作物科学和农业工程领域的进展，以开发稳健、可信且以农民为中心的决策支持系统。通过批判性评估技术创新和实施挑战，本综述提供了一个理解AI在农业中现状的框架，同时确定了将塑造下一代可持续和气候韧性农业系统的研究优先事项。","Discover Artificial Intelligence","2026-10-17T00:00:00Z",83,{"impact":18,"substance":18,"depth":19,"authority":66,"freshness":113,"relevant":22,"comment":195},"该综述系统梳理了AI在农业中的应用与落地瓶颈，覆盖精准灌溉、机器人、大模型决策支持等方向，信息密度高，但属综述类论文，产业级突破性有限，可入精选。",[197],{"name":191,"url":188},[118,119,29,199,200],"数字孪生","气候智慧农业",{"code":41,"slug":130,"name":131,"path":132,"parent_code":127},[],{"status":83,"locked":14,"version":36,"method":37,"evidence":204},{"candidates":205,"agricultural_evidence":45},[206,214,221,228,233,240,246],{"code":41,"score":207,"supported":45,"title":208,"abstract":210,"tags":213},19,[209],"artificial intelligence",[211,209,212,147,145],"计算机视觉","computer vision",[119,29],{"code":215,"score":53,"supported":45,"title":216,"abstract":217,"tags":220},"H2",[],[218,219],"土壤","soil",[],{"code":222,"score":53,"supported":45,"title":223,"abstract":224,"tags":227},"H3",[],[225,226],"灌溉","irrigation",[],{"code":150,"score":229,"supported":45,"title":230,"abstract":231,"tags":232},5,[],[199],[199,118],{"code":234,"score":235,"supported":45,"title":236,"abstract":237,"tags":239},"F6",3,[],[238],"农业机器人",[],{"code":241,"score":235,"supported":45,"title":242,"abstract":243,"tags":245},"G2",[],[244],"smallholder",[],{"code":76,"score":22,"supported":14,"title":247,"abstract":248,"tags":249},[],[],[200],[251,252],"气候智慧农业 AI","农业人工智能 气候智慧农业 大语言模型 数字孪生","气候智慧农业AI-5232","10.1007\u002Fs44163-026-02540-1",{"doi":254,"openalex_id":256,"authors":257,"venue":191,"cited_by_count":160,"oa_url":188,"card":261,"direction":181,"ingested_from":183},"W7222823256",[258],{"name":259,"orcid":260},"Mostafa Farajpour","https:\u002F\u002Forcid.org\u002F0000-0003-4223-502X",{"tldr":262,"method":263,"finding":264,"direction":265,"opportunity":266},"综述农业人工智能在精准灌溉、产量预测、病虫害监测等应用及落地挑战。","文献综述，覆盖视觉Transformer、基础模型、数字孪生、多模态AI与联邦学","农业AI落地受限于跨区域验证、数据碎片化与基础设施，需兼顾可解释性与公平治理。","农业人工智能与决策模型","可研究小农户场景下跨区域可泛化、可解释且低成本的农业大模型与联邦学习部署。","2026-10-18T23:30:07.666127Z",{"id":269,"title":270,"url":271,"summary":272,"summary_zh":273,"content":8,"source_name":274,"source_url":271,"published_at":275,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":160,"score_detail":276,"sources":278,"tags":280,"primary_topic":8,"secondary_topics":284,"classification":285,"search_phrases":299,"slug":302,"view_count":21,"doi":303,"paper":304,"created_at":341},3859,"Large language models for social governance: An analytical framework and research agenda","https:\u002F\u002Fdoi.org\u002F10.1093\u002Fnsr\u002Fnwag602","Abstract As social governance becomes increasingly complex, traditional governance approaches face challenges such as information asymmetry, fragmented coordination, delayed risk response, and uneven public service delivery. Large language models (LLMs) offer new possibilities for knowledge processing, decision support, and human–machine interaction in this context. This article provides a comprehensive and framework-oriented review of research on LLMs, digital governance, public governance, and computational social science. It develops a governance-oriented analytical framework that links LLMs capabilities, governance functions, operating mechanisms, governance outcomes, and risk constraints. The review examines representative application scenarios, including emergency management, information propagation, intelligent rule of law, data governance, public security, healthcare, and e-government. The analysis highlights uneven evidence maturity across these scenarios: bounded tasks such as policy retrieval, public consultation, information summarization, and administrative document processing have clearer deployment pathways, whereas high-stakes simulation, legal reasoning, public security assessment, healthcare triage, and cross-departmental decision support require stronger validation and oversight. LLM-enabled social governance also remains constrained by hallucination, algorithmic bias, privacy leakage, weak explainability, resource inequality, accountability gaps, and institutional adaptation challenges. This article argues that LLMs should be understood as decision-support infrastructures embedded in accountable, auditable, and public value-oriented governance institutions.","摘要 随着社会治理日益复杂，传统治理方式面临信息不对称、协调碎片化、风险响应滞后、公共服务供给不均等挑战。大语言模型（LLMs）为此情境下的知识处理、决策支持与人机交互提供了新的可能。本文对LLMs、数字治理、公共治理与计算社会科学的研究进行了综合性、框架导向的综述，构建了一个面向治理的分析框架，将LLMs能力、治理功能、运行机制、治理结果与风险约束相联结。综述考察了代表性应用场景，包括应急管理、信息传播、智慧法治、数据治理、公共安全、医疗健康与电子政务。分析表明，这些场景的证据成熟度不均衡：政策检索、公众咨询、信息摘要与行政文书处理等边界明确的任务具有更清晰的部署路径，而高风险模拟、法律推理、公共安全评估、医疗分诊与跨部门决策支持则需要更强的验证与监督。LLMs赋能的社会治理还受到幻觉、算法偏见、隐私泄露、可解释性薄弱、资源不平等、问责缺口与制度适应挑战等制约。本文认为，应将LLMs理解为嵌入可问责、可审计且面向公共价值的治理机构之中的决策支持基础设施。","National Science Review","2026-09-27T00:00:00Z",{"impact":160,"substance":160,"depth":160,"authority":160,"freshness":160,"relevant":160,"comment":277},"该文探讨大语言模型在社会治理中的应用框架，属公共管理与计算社会科学范畴，与三农、农业信息化、智慧农业、数字乡村等主题无直接关联，不建议进入每日精选。",[279],{"name":274,"url":271},[29,281,282,283],"数字治理","社会治理","公共管理",[],{"status":286,"locked":14,"version":36,"method":37,"evidence":287},"needs_review",{"candidates":288,"agricultural_evidence":14,"reason":298},[289,293],{"code":41,"score":67,"supported":45,"title":290,"abstract":291,"tags":292},[],[145],[29],{"code":294,"score":22,"supported":14,"title":295,"abstract":296,"tags":297},"G5",[],[],[281],"农业相关性需要复核",[300,301],"大语言模型 公共管理 数字治理 社会治理","大语言模型 公共管理","大语言模型公共管理数字治理社会治理-3859","10.1093\u002Fnsr\u002Fnwag602",{"doi":303,"openalex_id":305,"authors":306,"venue":274,"cited_by_count":160,"oa_url":334,"card":335,"direction":339,"ingested_from":183},"W7214589214",[307,309,312,313,315,317,318,321,324,326,327,328,331],{"name":308,"orcid":8},"Jun Hao",{"name":310,"orcid":311},"Jian Li","https:\u002F\u002Forcid.org\u002F0000-0002-4549-0694",{"name":310,"orcid":311},{"name":314,"orcid":8},"Linyuan Lü",{"name":316,"orcid":8},"Fang Wang",{"name":316,"orcid":8},{"name":319,"orcid":320},"Yijun Liu","https:\u002F\u002Forcid.org\u002F0009-0008-6601-9237",{"name":322,"orcid":323},"Jianjun Yu","https:\u002F\u002Forcid.org\u002F0000-0002-5203-171X",{"name":325,"orcid":8},"Changyong Liang",{"name":322,"orcid":323},{"name":325,"orcid":8},{"name":329,"orcid":330},"Xiaolong Zheng","https:\u002F\u002Forcid.org\u002F0000-0003-0405-5458",{"name":332,"orcid":333},"Zhixin Liu","https:\u002F\u002Forcid.org\u002F0000-0003-2342-4298","https:\u002F\u002Facademic.oup.com\u002Fnsr\u002Fadvance-article-pdf\u002Fdoi\u002F10.1093\u002Fnsr\u002Fnwag602\u002F71450964\u002Fnwag602.pdf",{"tldr":336,"method":337,"finding":338,"direction":339,"opportunity":340},"综述大语言模型在社会治理中的应用，构建分析框架并提出研究议程。","文献综述，整合LLM、数字治理与计算社会科学研究。","低风险任务部署路径清晰，高风险场景需验证与监管，LLM应作为决策支持基础设施。","数字乡村与农业信息化","可探索LLM在乡村治理、农业政策咨询等低风险场景的落地与评估框架。","2026-09-30T23:30:46.382519Z",{"id":343,"title":344,"url":345,"summary":346,"summary_zh":347,"content":8,"source_name":348,"source_url":345,"published_at":349,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":350,"score_detail":351,"sources":353,"tags":355,"primary_topic":358,"secondary_topics":364,"classification":365,"search_phrases":395,"slug":398,"view_count":22,"doi":399,"paper":400,"created_at":411},3665,"ASAS-NANP SYMPOSIUM: MATHEMATICAL MODELING IN ANIMAL NUTRITION: The Evolution of Large Language Models and Their Impact on Animal Sciences","https:\u002F\u002Fdoi.org\u002F10.1093\u002Fjas\u002Fskag306","The rapid rise of large language models (LLM) is reshaping the scientific landscape, transitioning from early statistical language models to advanced transformer-based architectures capable of synthesizing knowledge across disciplines. While their predictive capacity and scalability have opened new avenues in data analysis, hypothesis generation, and decision support, concerns remain regarding bias, hallucination, reproducibility, and ethical governance. In animal sciences, LLM are gradually applied to challenges in nutrition modeling, animal health, genetic selection, and sustainability. Precision nutrition has benefited from LLM-driven synthesis of feed and metabolic data, enabling individualized feeding strategies and improved resource efficiency. In animal health, domain-specific systems have demonstrated applications in diagnostics and epidemiological monitoring. LLM is augmenting genomic analyses to accelerate marker discovery and breeding optimization, while sustainability efforts employ them to model greenhouse gas emissions, feed additives, and adaptation to climatic stressors. Notably, decision-support platforms demonstrate how domain-specialized LLM can bridge mechanistic knowledge with predictive analytics, enhancing knowledge transfer and empowering livestock producers. However, risks associated with overreliance, recursive reuse of LLM outputs in model development, and pseudo-expertise underscore the critical importance of human oversight. Unlike mechanistic models, which embed biological causality, LLM are entirely data-driven and may confidently propagate errors if trained on ill-conditioned datasets. Responsible use requires transparent reporting, validation, and bias auditing, with domain-specific fine-tuning. Open-source models can enhance reproducibility and trust, but they also raise financial and security concerns. LLM must be guided by transparency, accountability, and fairness to ensure they complement, rather than replace, human expertise. By advancing inquiry and livestock management, LLM hold the potential to support sustainable food production systems if deployed responsibly. Rather than full training, most applications will rely on fine-tuning and augmentation, which are more sustainable and adaptable strategies. This review intends to promote critical literacy and responsible adoption of LLM in research, education, and livestock decision-support systems for animal scientists, nutrition modelers, and extension professionals.","大语言模型（LLM）的迅速崛起正在重塑科学格局，从早期的统计语言模型过渡到能够跨学科综合知识的先进基于Transformer的架构。尽管其预测能力和可扩展性在数据分析、假设生成和决策支持方面开辟了新途径，但围绕偏见、幻觉、可重复性和伦理治理的担忧依然存在。在动物科学领域，LLM正逐步应用于营养建模、动物健康、遗传选择和可持续性等方面的挑战。精准营养已受益于LLM驱动的饲料与代谢数据综合，从而实现个体化饲喂策略并提高资源效率。在动物健康方面，领域专用系统已在诊断和流行病学监测中展现出应用价值。LLM正在增强基因组分析，以加速标记发现和育种优化，而可持续性工作则利用其建模温室气体排放、饲料添加剂及对气候胁迫的适应性。值得注意的是，决策支持平台展示了领域专用LLM如何将机制性知识与预测分析相衔接，促进知识转化并赋能畜牧生产者。然而，过度依赖、在模型开发中递归复用LLM输出以及伪专业知识等风险，凸显了人类监督的至关重要性。与嵌入生物学因果关系的机制性模型不同，LLM完全由数据驱动，若在病态数据集上训练，可能自信地传播错误。负责任的使用需要透明报告、验证和偏见审计，并进行领域专用微调。开源模型可以增强可重复性和信任，但也引发财务和安全方面的担忧。LLM必须以透明性、问责制和公平性为指引，以确保其补充而非取代人类专业知识。通过在研究和畜牧管理中推动探究，LLM在负责任部署的前提下具有支持可持续食品生产系统的潜力。大多数应用将依赖微调和增强，而非完整训练，这是更具可持续性和适应性的策略。本综述旨在促进动物科学家、营养建模人员和推广专业人员在研究、教育和畜牧决策支持系统中对LLM的批判性素养和负责任采用。","Journal of Animal Science","2026-09-25T00:00:00Z",78,{"impact":19,"substance":17,"depth":19,"authority":20,"freshness":100,"relevant":22,"comment":352},"核心期刊综述，系统梳理大语言模型在动物营养、健康与育种中的应用与风险，专业深度高但属综述性论文，产业级影响有限。",[354],{"name":348,"url":345},[118,119,29,356,357],"精准营养","动物科学",{"code":359,"slug":360,"name":361,"path":362,"parent_code":363},"B7","protected-horticulture","设施园艺与植物工厂","\u002Ftopic\u002Fhorticulture\u002Fprotected-horticulture","B",[],{"status":83,"locked":14,"version":36,"method":37,"evidence":366},{"candidates":367,"agricultural_evidence":45},[368,374,381,385,391],{"code":359,"score":53,"supported":45,"title":369,"abstract":370,"tags":373},[],[371,372],"温室","greenhouse",[],{"code":375,"score":53,"supported":45,"title":376,"abstract":377,"tags":380},"C",[],[378,379],"畜牧","livestock",[],{"code":41,"score":229,"supported":45,"title":382,"abstract":383,"tags":384},[],[145],[119,29],{"code":386,"score":235,"supported":45,"title":387,"abstract":388,"tags":390},"C4",[],[389],"饲料",[],{"code":150,"score":22,"supported":14,"title":392,"abstract":393,"tags":394},[],[],[118],[396,397],"大语言模型 动物科学","LLM 精准营养","大语言模型动物科学-3665","10.1093\u002Fjas\u002Fskag306",{"doi":399,"openalex_id":401,"authors":402,"venue":348,"cited_by_count":160,"oa_url":405,"card":406,"direction":181,"ingested_from":183},"W7214347972",[403],{"name":404,"orcid":8},"Luis O Tedeschi","https:\u002F\u002Facademic.oup.com\u002Fjas\u002Fadvance-article-pdf\u002Fdoi\u002F10.1093\u002Fjas\u002Fskag306\u002F71398322\u002Fskag306.pdf",{"tldr":407,"method":408,"finding":409,"direction":265,"opportunity":410},"综述大语言模型在动物科学中的应用、风险与负责任使用路径。","文献综述，梳理LLM在营养、健康、遗传和可持续性中的应用。","LLM可辅助精准营养与决策支持，但需人类监督、验证和领域微调。","可探索面向畜牧生产的领域专用LLM微调与可验证决策支持系统，并建立偏差审计与可复现评估框架。","2026-09-28T23:30:12.831387Z",{"id":413,"title":414,"url":415,"summary":416,"summary_zh":8,"content":417,"source_name":418,"source_url":8,"published_at":419,"category":420,"cover_url":8,"hotness":13,"is_selected":45,"score":421,"score_detail":422,"sources":426,"tags":428,"primary_topic":8,"secondary_topics":432,"classification":433,"search_phrases":456,"slug":459,"view_count":151,"doi":8,"paper":8,"created_at":460},3596,"中国农科院基因组所商连光团队开发自进化AI科学家'绿洲'助力作物耐盐碱研究——Molecular Plant在线发表","https:\u002F\u002Fagis.caas.cn\u002Fxwzx\u002Fkyjz\u002Fb0960d967c964346b50cb50c4886c209.htm","2026年9月18日，中国农科院基因组所（大鹏湾实验室）商连光团队联合崖州湾国家实验室、盐碱地综合利用国家技术创新中心，在《分子植物（Molecular Plant）》上在线发表题为'OASIS, a self-evolving AI scientist that integrates omics data and literature knowledge for plant stress research'的研究论文。该研究开发了面向植物逆境研究的自进化AI科学家'绿洲'（OASIS）。系统将大语言模型、多智能体协作、植物逆境文献知识库以及组学和遗传数据纳入同一科研流程。'绿洲'将复杂科研任务拆解为多个子任务，由6个AI智能体分工协作完成。研究团队进一步以水稻耐盐为案例，对'绿洲'提出的候选基因开展实验验证，形成了从科学问题到候选假说再到实验验证的完整研究路径。","2026年9月18日，中国农科院基因组所（大鹏湾实验室）商连光团队联合崖州湾国家实验室、盐碱地综合利用国家技术创新中心，在《分子植物（Molecular Plant）》上在线发表题为“OASIS, a self-evolving AI scientist that integrates omics data and literature knowledge for plant stress research”的研究论文。\n\n![Image 1](https:\u002F\u002Fagis.caas.cn\u002Fimages\u002F2026-09\u002Fefa2de3e737741ea8e2ae4ceeed0695f.png)\n\n该研究开发了面向植物逆境研究的自进化AI科学家“绿洲”（OASIS）。系统将大语言模型、多智能体协作、植物逆境文献知识库以及组学和遗传数据纳入同一科研流程，可围绕一个科学问题自主完成任务规划、文献检索、数据查询、文献与组学数据融合推理和结果审查，并能够从历史任务中的失败与纠错中学习可复用经验。研究团队进一步以水稻耐盐为案例，对“绿洲”提出的候选基因开展实验验证，形成了从科学问题到候选假说再到实验验证的完整研究路径。\n\n植物抗逆研究往往不是“缺少信息”，而是证据分散在不同载体中：机制研究主要沉淀在论文里，基因表达、遗传定位、群体变异和种质资源则分布在多个数据库中。研究人员通常需要在大量文献和多类数据之间反复检索、比对和判断，才能把一个科学问题逐步收敛为可供实验验证的假说。通用大语言模型擅长理解和生成文本，但仅依靠模型自身知识，仍难以稳定完成跨数据库、跨数据类型的连续科研分析。\n\n**从“读文献”到“查数据”，让AI进入真实科研工作流**\n\n针对这一问题，“绿洲”将复杂科研任务拆解为多个子任务，由6个AI智能体分工协作完成。系统先识别研究意图并制定任务计划，再根据需要调用文献检索、数据分析和科研数据库等工具，最终由审查与综合模块检查证据覆盖范围和任务完成情况后生成结论。任务规划、文献片段、数据库查询结果和分析记录都会保存在共享工作空间中，便于研究人员回溯和核对。整个过程类似于一个小型科研团队围绕同一科学问题持续协作。\n\n![Image 2](https:\u002F\u002Fagis.caas.cn\u002Fimages\u002F2026-09\u002Ff4ecb718f8d942b89d22b99880d0dbe3.png)\n\n图1 | “绿洲”多智能体协作科研框架。系统围绕科学问题进行任务拆解，并协调文献检索、数据分析、结果审查和综合生成，使不同来源的科研证据能够在同一工作流中被连续调用和融合推理。\n\n支撑这一流程的是专门整理的植物逆境知识和数据资源。“绿洲”目前覆盖拟南芥、水稻、玉米、小麦、大豆和陆地棉6个物种，以及盐、干旱、高温和低温4类主要非生物胁迫。文献知识库收录超过15万篇相关论文摘要和1.5万篇开放获取全文，并持续同步新发表研究；同时接入植物逆境表达数据、GWAS\u002FQTL、群体变异、eQTL和种质资源等结构化信息，使系统不仅能够回答“文献中有哪些报道”，还能够进一步判断“数据是否支持这一推断”。\n\n**OASIS的优势，体现在数据驱动的研究任务中**\n\n为评估系统在真实科研任务中的表现，研究团队构建了面向植物逆境研究的基准测试集PlantStressQA，包含200个贴近真实科研过程的问题，覆盖机制证据、基因ID映射、表达分析、遗传位点分析、品种筛选和群体变异六类任务。\n\n在与多种通用大语言模型的比较中，“绿洲”总体得分为84.0分，参评通用模型得分为33.8至50.1分。在需要实际调用并整合结构化数据的任务中，“绿洲”优势明显，其中群体变异和遗传位点分析两类任务分别比表现最好的通用模型高55.6分和52.3分，在基因ID映射和表达证据分析中也表现出明显优势。\n\n![Image 3](https:\u002F\u002Fagis.caas.cn\u002Fimages\u002F2026-09\u002Fc2b2782477a64a55a3366b44219320d5.png)\n\n图2 | “绿洲”在PlantStressQA上的系统评价。系统在基因ID转换、表达数据查询、遗传位点分析和群体变异分析等数据密集型任务中表现出更明显的优势。\n\n这些结果说明，“绿洲”的核心并不是单纯依赖语言模型“记住更多知识”，而是通过组织科研工具和数据资源，在研究过程中主动查找数据、完成格式转换、比较不同来源的证据，并将这些步骤串联为可追溯的分析过程。\n\n**做错一次之后，下一次还会不会犯同样的错误？**\n\n真实科研分析中，数据库字段不匹配、基因ID格式错误、工具调用顺序不合理等问题并不少见。如果系统每次遇到相似任务都从头试错，不仅效率低，也会造成额外的计算资源消耗。为此，“绿洲”设计了自进化经验学习模块（SEEL），从历史任务中识别“先失败、随后被成功纠正”的执行轨迹，对比错误路径和正确路径，提炼可复用的操作经验，并通过重新执行真实任务检验其有效性。只有验证通过的经验才会进入技能库，供后续相似任务调用。\n\n![Image 4](https:\u002F\u002Fagis.caas.cn\u002Fimages\u002F2026-09\u002F72cd8d054dc14f94b7b9461f4a7f8c46.png)\n\n图3 | SEEL自进化经验学习流程及效果。系统从历史任务的失败与纠错过程中提炼可复用经验，经验证后纳入技能库。\n\n研究团队从PlantStressQA的历史轨迹中筛选出86个包含可复用经验的任务。应用这些经验后，工具调用失败、数据库错误、重复操作和模型处理文本量均有所减少。在一个表达分析任务中，大模型调用次数从75次降至42次，模型处理文本量从112万降至31万，运行时间从476秒缩短到236秒，数据库错误和失败工具调用则从多次降为0次。随着有效经验持续积累，“绿洲”能够在后续相似任务中减少重复试错并优化执行过程。\n\n**最关键的一步：AI提出的候选基因，实验能不能验证？**\n\n基准测试可以衡量AI系统是否“会做题”，但科研价值最终还要回到能否帮助提出值得验证的科学假说。研究团队以水稻耐盐为案例，在未预设具体候选基因的情况下，通过多轮提问引导“绿洲”开展分析。系统先从已有研究中归纳离子稳态与Na⁺\u002FK⁺运输、ABA信号与渗透调节、ROS稳态、MAPK级联和盐感知\u002FCa²⁺信号等主要耐盐调控模块，并整理出284个有文献证据支持的拟南芥调控基因；随后开展跨物种同源映射和功能筛选，再整合水稻耐盐GWAS\u002FQTL等遗传证据，最终逐步确定5个优先候选基因。\n\n![Image 5](https:\u002F\u002Fagis.caas.cn\u002Fimages\u002F2026-09\u002Fa9ba4a0ff48e43bc8239834881fb034c.png)\n\n图4 | “绿洲”辅助筛选水稻耐盐候选基因的多轮对话示例。系统整合耐盐机制文献、水稻同源基因信息和群体遗传证据，逐步筛选出5个优先候选基因。\n\n研究团队根据“绿洲”给出的证据链，选择此前未被直接报道参与水稻耐盐调控的OsPP2a（LOC_Os12g04290）开展实验验证。该基因的拟南芥同源基因TAP46与ABA信号相关，OsPP2a本身位于水稻耐盐meta-QTL区间内，并与一个eQTL信号共定位。两个独立的Ospp2a突变材料在150 mM NaCl处理后均表现出更明显的盐敏感表型，恢复存活率降低，同时Na⁺积累增加、Na⁺\u002FK⁺平衡发生改变。RT-qPCR结果还显示，离子稳态相关基因OsSOS1和OsHKT1;5的表达发生变化，为OsPP2a参与水稻盐胁迫响应提供了遗传和分子层面的证据。\n\n![Image 6](https:\u002F\u002Fagis.caas.cn\u002Fimages\u002F2026-09\u002Fe698f656e5c8488fb80baadceaab1bf0.png)\n\n图5 | “绿洲”优先推荐基因OsPP2a的耐盐功能验证。两个独立突变材料在盐胁迫下表现出更低的恢复存活率和更高的Na⁺\u002FK⁺比值，支持OsPP2a参与水稻耐盐响应。\n\n在这一案例中，研究人员通过多轮提问引导“绿洲”完成机制梳理、跨物种证据迁移、遗传数据查询和候选基因优先级排序，再由实验进一步检验候选假说，形成了“科学问题→证据融合→候选假说→实验验证”相衔接的研究路径。\n\n**从“AI回答科学”到“AI参与科学”**\n\n“绿洲”的目标并不是用AI替代科研人员作出判断，而是把过去需要在多篇文献、多个数据库和不同分析工具之间往返完成的工作，组织成一条可追溯、可复核的科研流程。通过多智能体协作、文献与组学数据融合推理以及自进化经验学习，系统能够帮助研究人员更高效地汇集证据、缩小候选范围并提出更有依据的科学假说。\n\n这项工作为人工智能参与植物抗逆研究提供了一个从“回答问题”走向“参与分析”的实践框架。未来，随着文献、组学数据和科研工具持续扩展，“绿洲”有望进一步服务于作物抗逆机制解析、关键基因挖掘和后续实验研究，为耐盐碱等作物抗逆改良提供智能化研究工具。\n\n基因组所（大鹏湾实验室）商连光研究员和崖州湾国家实验室钱前院士为论文共同通讯作者，基因组所（大鹏湾实验室）博士后韩漾和副研究员魏华为论文共同第一作者。该研究得到国家重点研发计划和国家自然科学基金的资助。\n\n平台入口：[https:\u002F\u002Fwww.oasis.ac.cn](https:\u002F\u002Fwww.oasis.ac.cn\u002F)\n\n原文链接：[https:\u002F\u002Fwww.cell.com\u002Fmolecular-plant\u002Fabstract\u002FS1674-2052(26)00305-9](https:\u002F\u002Fwww.cell.com\u002Fmolecular-plant\u002Fabstract\u002FS1674-2052(26)00305-9)","中国农科院基因组研究所 2026-09-18","2026-09-18T00:00:00Z","报道",90,{"impact":423,"substance":424,"depth":207,"authority":65,"freshness":53,"relevant":22,"comment":425},26,24,"国家级科研机构在《分子植物》发表的自进化AI科学家系统，实现文献与组学数据融合推理并经水稻耐盐基因实验验证，属农业人工智能与智慧育种领域的标志性突破。",[427],{"name":418,"url":415},[119,429,30,430,431],"水稻","耐盐碱","智慧育种",[],{"status":286,"locked":14,"version":36,"method":37,"evidence":434},{"candidates":435,"agricultural_evidence":45,"reason":455},[436,441,446,450],{"code":437,"score":67,"supported":45,"title":438,"abstract":439,"tags":440},"A1",[],[429],[429],{"code":215,"score":67,"supported":45,"title":442,"abstract":443,"tags":445},[],[444],"盐碱地",[430],{"code":41,"score":21,"supported":14,"title":447,"abstract":448,"tags":449},[],[],[119,30],{"code":451,"score":22,"supported":14,"title":452,"abstract":453,"tags":454},"F2",[],[],[431],"多个产业证据相当，需要确认主归属",[457,458],"中国农科院基因组所 绿洲 OASIS","商连光 水稻耐盐 OsPP2a","中国农科院基因组所绿洲OASIS-3596","2026-09-27T00:05:16.998310Z"]