{ "claimSet": { "claims": [ { "boundary": "共识来自立场论文和研讨会报告，缺少对机制效果的定量验证；不同学科对开放共享的接受度和可行性存在差异。", "claimType": "cross_paper_synthesis", "contradictingPaperIds": [], "id": "C1", "maturity": "supported", "statement": "AI for Science 的核心共识是，其成功不仅依赖算法改进，还取决于跨学科社区建设、数据基础设施和开放科学机制；单纯方法突破不足以解决科学采用障碍。", "supportingPaperIds": [ "openalex:W4323697696", "openalex:W4401726605", "arxiv:2509.02661", "arxiv:2310.18852" ], "whyItMatters": "这改变了将 AI for Science 视为纯技术问题的认知，指出组织与制度因素同样是瓶颈。" }, { "boundary": "主要证据来自材料科学和生物化学综述，跨领域普适性尚未经过系统验证；数据充分性的定量阈值不存在。", "claimType": "cross_paper_synthesis", "contradictingPaperIds": [], "id": "C2", "maturity": "supported", "statement": "在材料科学和生物化学等领域，高质量、大规模且标准化的数据是 AI 模型性能的主要限制因素；仅遵循 FAIR 原则不足以保证 AI 就绪性，还需要评估数据充分性。", "supportingPaperIds": [ "openalex:W4402901051", "openalex:W4406828533" ], "whyItMatters": "指导资源投入优先解决数据准备问题，而不是盲目追求更大模型。" }, { "boundary": "主要基于立场论文和案例，未证明在所有科学任务中必须人类介入；ARU 标准本身尚未在大规模实践中检验。", "claimType": "cross_paper_synthesis", "contradictingPaperIds": [], "id": "C3", "maturity": "supported", "statement": "多个分支强调可解释性和人类专家介入对科学发现至关重要；仅靠事后解释不足以保证科学结论，需要结合准确性、可再现性和可理解性标准，并用实验进行裁决。", "supportingPaperIds": [ "arxiv:2406.10557", "openalex:W3205208140", "arxiv:2509.02661" ], "whyItMatters": "防止把 AI 输出当作科学真理，推动建立可解释性和验证规范。" }, { "boundary": "证据多为案例报告和综述二手信息，缺少系统比较和失败案例；闭环验证多在数字环境，物理实验自动化仍在路线图中。", "claimType": "cross_paper_synthesis", "contradictingPaperIds": [], "id": "C4", "maturity": "supported", "statement": "AI for Science 正在从单一预测模型向自主发现、闭环实验的‘代理科学’演进，但当前成功案例（如虚拟实验室、机器集体智能）主要是在受控基准或数字阶段的存在性证明，尚未形成可推广到物理实验的成熟能力。", "supportingPaperIds": [ "arxiv:2508.14111", "arxiv:2604.27297", "arxiv:2608.02775", "arxiv:2607.09025" ], "whyItMatters": "帮助区分宣传与已验证能力，避免高估自主科学 AI 的现状。" }, { "boundary": "结论基于单一综述的案例整理，未做跨方法定量比较；是否能替代传统求解器无定论，正面案例可能带有选择偏差。", "claimType": "direct_finding", "contradictingPaperIds": [], "id": "C5", "maturity": "single_source", "statement": "物理信息神经网络（PINNs）通过将偏微分方程嵌入损失函数，在流体、固体力学等领域展现出作为数值求解器替代或补充的潜力，但在高维问题、可扩展性和可靠性方面仍存在显著未解决挑战。", "supportingPaperIds": [ "openalex:W4412642369" ], "whyItMatters": "为科学计算研究者提供方法选择参考，并指出 PINN 尚未成熟的边界。" }, { "boundary": "该综述仅覆盖生物/化学 Transformer 模型，排除图神经网络、扩散模型；其结论反映 2024 年状态，可能已过时。", "claimType": "direct_finding", "contradictingPaperIds": [], "id": "C6", "maturity": "single_source", "statement": "生物与化学科学大语言模型已发展出文本、分子、蛋白质、基因组和多模态五类，但评估基础设施滞后：缺乏大学后水平基准，且计算指标不能直接代表湿实验有效性。", "supportingPaperIds": [ "openalex:W4406828533" ], "whyItMatters": "提醒研究者不要仅凭基准分数判断科学 LLM 的实用价值，推动开发更贴近实验的评测。" }, { "boundary": "基于 2021 年发表时的少数典型案例，不能量化错误率；随着版本更新，某些限制可能已改善，但模型缺乏能量最小化、无法涵盖翻译后修饰等边界可能仍然存在。", "claimType": "direct_finding", "contradictingPaperIds": [], "id": "C7", "maturity": "single_source", "statement": "AlphaFold2 的高置信度预测并不等于生物学上正确的结构；部分预测（如 CENP-E、Mad2）与实验结构不符，说明实验结构生物学仍是判定生物学真实状态和构象的必要手段。", "supportingPaperIds": [ "openalex:W3205208140" ], "whyItMatters": "纠正对 AI 蛋白结构预测的过度信任，支持继续投资实验基础设施和模型作为假设生成器的使用方式。" }, { "boundary": "基准仅使用 1000 样本、加性同方差噪声和静态因果图，非线性类型有限；不能代表所有真实时间序列因果场景。", "claimType": "direct_finding", "contradictingPaperIds": [], "id": "C8", "maturity": "single_source", "statement": "因果发现基准实验表明，时间序列因果推断方法（如 Granger、PCMCI、DYNOTEARS）的性能对因果充分性、线性、无瞬时效应等假设高度敏感；违反假设时 F1 显著下降，且超参数选择（如 PCMCI 的 p 值阈值）对结果影响强烈。", "supportingPaperIds": [ "arxiv:2104.08043" ], "whyItMatters": "为因果发现方法选择提供实证依据，强调必须检查假设并进行超参数调优，避免盲目应用。" }, { "boundary": "结果局限于所选数据集和超参数配置，能耗为事后估计而非实测，未做统计显著性检验，因此排名差异可能是探索性的。", "claimType": "direct_finding", "contradictingPaperIds": [], "id": "C9", "maturity": "single_source", "statement": "更新版符号回归基准（SRBench）显示，没有任何单一符号回归算法在所有数据集上全面领先；能耗、超参数搜索空间和数据噪声对算法排名有实质性影响，且高复杂度模型可能只是拟合噪声。", "supportingPaperIds": [ "arxiv:2505.03977" ], "whyItMatters": "推动符号回归研究从追求单一指标转向考虑资源效率和鲁棒性，为算法选型提供更现实依据。" }, { "boundary": "结论基于对现有技术的综述和单一案例（AlphaFold2），并未量化科学领域内存技术未充分利用的程度，推广到其他科学架构需谨慎。", "claimType": "direct_finding", "contradictingPaperIds": [], "id": "C10", "maturity": "single_source", "statement": "科学 Transformer 模型（如 AlphaFold2）的训练受‘AI 内存墙’制约，现有通用内存优化技术未充分应用于科学架构，需要针对 Evoformer、SE(3)-Transformer 等定制优化，如 AlphaFold2 中的序列维动态并行。", "supportingPaperIds": [ "arxiv:2501.11847" ], "whyItMatters": "指出科学大模型扩展的计算基础设施瓶颈，为工程优化提供方向。" }, { "boundary": "仅限于大语言模型，表格对比具有时效性且依赖模型版本；对象方向研究多为检测分类，缺乏机制解释。", "claimType": "direct_finding", "contradictingPaperIds": [], "id": "C11", "maturity": "single_source", "statement": "在社会科学中，将 AI 作为工具在速度、成本、可重复性和可扩展性方面通常优于传统方法，但在效度、数值分析和机制解释方面较弱；LLM 能否替代人类受试者仍是开放问题，目前不能取代社会科学家。", "supportingPaperIds": [ "arxiv:2401.11839" ], "whyItMatters": "为社会科学研究采用 AI 提供权衡框架，避免仅因效率而牺牲效度。" }, { "boundary": "该框架是分析性工具而非可运行算法，其关于累积洞察力的主张是逻辑推断而非实证结果；社区级共享依赖尚未解决激励机制。", "claimType": "author_claim", "contradictingPaperIds": [], "id": "C12", "maturity": "single_source", "statement": "演化智能视角认为，累积科学发现的关键不仅在于候选解优化，还在于保留搜索历史和失败候选；失败实验和谱系记录可以作为科学证据，用于提炼设计规则和可测试假设。", "supportingPaperIds": [ "arxiv:2607.09025" ], "whyItMatters": "改变对失败的认知，推动构建保留完整搜索过程的基础设施，提高研究可追溯性。" } ], "evidenceMode": "fulltext", "schemaVersion": 1 }, "editorialPlan": { "centralThesis": "围绕“AI for Science：领域全景与核心共识：关于 AI for Science，现有研究形成了哪些较可信且容易理解的核心结论？主要有哪些研究分支和代表性证据？”，应先区分当前证据直接支持的结论与仍待验证的推断。", "evidenceMode": "fulltext", "modules": [ { "argumentRole": "orient", "avoidRepeatingClaimIds": [], "claimIds": [], "confidencePolicy": "evidence_calibrated", "confusionToResolve": "", "evidencePaperIds": [ "openalex:W4401726605", "openalex:W4412642369", "openalex:W4406828533" ], "exampleRequirement": "none", "id": "M1", "includeReason": "先建立读者理解后续结论所需的共同语境。", "kind": "orientation", "lengthBudget": 300, "readerQuestion": "这项研究问题的范围和阅读入口是什么？", "readerTakeaway": "先明确问题范围和阅读入口。", "renderMode": "prose", "requirements": [ "说明问题边界", "避免把研究背景写成结论" ], "title": "如何理解这个问题", "transitionFromPrevious": "开篇建立共同语境。" }, { "argumentRole": "answer", "avoidRepeatingClaimIds": [], "claimIds": [ "C1", "C2", "C3", "C4", "C5", "C6", "C7", "C8" ], "confidencePolicy": "evidence_calibrated", "confusionToResolve": "", "evidencePaperIds": [ "openalex:W4401726605", "openalex:W4412642369", "openalex:W4406828533", "arxiv:2604.27297", "arxiv:2508.14111", "arxiv:2509.02661", "openalex:W3205208140", "openalex:W4323697696", "arxiv:2607.09025", "arxiv:2310.18852", "arxiv:2406.10557", "arxiv:2401.11839", "arxiv:2501.11847", "arxiv:2608.02775", "openalex:W4402901051", "arxiv:2110.01831", "arxiv:2505.03977", "arxiv:2603.28361", "arxiv:2104.08043" ], "exampleRequirement": "concrete_example", "id": "M2", "includeReason": "让读者先获得能够独立理解的结论。", "kind": "core_conclusions", "lengthBudget": 1100, "readerQuestion": "当前证据最直接支持哪些结论？", "readerTakeaway": "读者能够复述当前证据支持的核心认识。", "renderMode": "prose", "requirements": [ "每条结论说明重要性", "结论与证据强度相匹配" ], "title": "目前可以带走的核心结论", "transitionFromPrevious": "在问题定向后直接回答研究问题。" }, { "argumentRole": "synthesize", "avoidRepeatingClaimIds": [], "claimIds": [ "C1", "C2", "C3", "C4", "C5", "C6", "C7", "C8", "C9", "C10", "C11", "C12" ], "confidencePolicy": "evidence_calibrated", "confusionToResolve": "", "evidencePaperIds": [ "openalex:W4401726605", "openalex:W4412642369", "openalex:W4406828533", "arxiv:2604.27297", "arxiv:2508.14111", "arxiv:2509.02661", "openalex:W3205208140", "openalex:W4323697696", "arxiv:2607.09025", "arxiv:2310.18852", "arxiv:2406.10557", "arxiv:2401.11839", "arxiv:2501.11847", "arxiv:2608.02775", "openalex:W4402901051", "arxiv:2110.01831", "arxiv:2505.03977", "arxiv:2603.28361", "arxiv:2104.08043" ], "exampleRequirement": "contrast_pair", "id": "M3", "includeReason": "帮助读者理解不同工作之间的关系。", "kind": "research_landscape", "lengthBudget": 900, "readerQuestion": "现有研究主要从哪些问题入口展开？", "readerTakeaway": "读者能够理解不同工作围绕哪些问题形成分支。", "renderMode": "map", "requirements": [ "按问题而不是论文顺序组织", "说明各分支之间的关系" ], "title": "当前研究版图", "transitionFromPrevious": "核心结论之后解释这些认识在研究版图中的关系。" }, { "argumentRole": "assess_evidence", "avoidRepeatingClaimIds": [], "claimIds": [ "C1", "C2", "C3", "C4", "C5", "C6", "C7", "C8", "C9", "C10", "C11", "C12" ], "confidencePolicy": "evidence_calibrated", "confusionToResolve": "", "evidencePaperIds": [ "openalex:W4401726605", "openalex:W4412642369", "openalex:W4406828533", "arxiv:2604.27297", "arxiv:2508.14111", "arxiv:2509.02661", "openalex:W3205208140", "openalex:W4323697696", "arxiv:2607.09025", "arxiv:2310.18852", "arxiv:2406.10557", "arxiv:2401.11839", "arxiv:2501.11847", "arxiv:2608.02775", "openalex:W4402901051", "arxiv:2110.01831", "arxiv:2505.03977", "arxiv:2603.28361", "arxiv:2104.08043" ], "exampleRequirement": "none", "id": "M4", "includeReason": "防止把有限证据写成领域共识。", "kind": "evidence_boundaries", "lengthBudget": 700, "readerQuestion": "当前证据没有回答什么？", "readerTakeaway": "读者能够区分已获支持的判断与仍待验证的推断。", "renderMode": "prose", "requirements": [ "区分缺失信息和反对证据", "明确仍需全文或新研究验证的部分" ], "title": "这些结论能相信到什么程度", "transitionFromPrevious": "在全文结尾校准前述判断的适用范围。" } ], "narrativeArc": [ "先明确问题范围和阅读入口。", "读者能够复述当前证据支持的核心认识。", "读者能够理解不同工作围绕哪些问题形成分支。", "读者能够区分已获支持的判断与仍待验证的推断。" ], "omittedModules": [ { "kind": "method_evolution", "reason": "未在规划阶段确认足够清晰的问题—方法—代价演进链。" }, { "kind": "system_layers", "reason": "未在规划阶段确认稳定的系统层级关系。" } ], "readerTakeaways": [ "AI for Science 的核心共识是，其成功不仅依赖算法改进，还取决于跨学科社区建设、数据基础设施和开放科学机制；单纯方法突破不足以解决科学采用障碍。", "在材料科学和生物化学等领域，高质量、大规模且标准化的数据是 AI 模型性能的主要限制因素；仅遵循 FAIR 原则不足以保证 AI 就绪性，还需要评估数据充分性。", "多个分支强调可解释性和人类专家介入对科学发现至关重要；仅靠事后解释不足以保证科学结论，需要结合准确性、可再现性和可理解性标准，并用实验进行裁决。", "AI for Science 正在从单一预测模型向自主发现、闭环实验的‘代理科学’演进，但当前成功案例（如虚拟实验室、机器集体智能）主要是在受控基准或数字阶段的存在性证明，尚未形成可推广到物理实验的成熟能力。", "物理信息神经网络（PINNs）通过将偏微分方程嵌入损失函数，在流体、固体力学等领域展现出作为数值求解器替代或补充的潜力，但在高维问题、可扩展性和可靠性方面仍存在显著未解决挑战。", "生物与化学科学大语言模型已发展出文本、分子、蛋白质、基因组和多模态五类，但评估基础设施滞后：缺乏大学后水平基准，且计算指标不能直接代表湿实验有效性。", "AlphaFold2 的高置信度预测并不等于生物学上正确的结构；部分预测（如 CENP-E、Mad2）与实验结构不符，说明实验结构生物学仍是判定生物学真实状态和构象的必要手段。", "因果发现基准实验表明，时间序列因果推断方法（如 Granger、PCMCI、DYNOTEARS）的性能对因果充分性、线性、无瞬时效应等假设高度敏感；违反假设时 F1 显著下降，且超参数选择（如 PCMCI 的 p 值阈值）对结果影响强烈。" ], "schemaVersion": 2 }, "insights": [ { "boundary": "证据来自立场论文、研讨会报告和形式化框架，暂无受控比较或干预研究；不同学科对开放共享的激励和风险结构不同，不能将材料科学或高能物理的共识直接套用到临床数据等强隐私领域。", "claimId": "C1", "confidence": "supported", "evidencePaperIds": [ "openalex:W4323697696", "openalex:W4401726605", "arxiv:2509.02661", "arxiv:2310.18852" ], "explanation": "Lawrence & Montgomery、Berens 等、AI+MPS 和 AI4OS 从不同入口汇聚到同一判断：前者把科学采用类比为技术扩散，指出 COVID-19 中英国 AI 几乎未进入实际响应；Berens 等将 AI for Science 定义为需要研究者的交汇点；AI+MPS 把社区、数据与教育列为三大支柱；AI4OS 要求把数据上下文和先验知识作为团队间传递的产证。差异在于制度落点：有的是国家层面资金与培训，有的是开放科学伦理；共同点是均未提供定量效果检验。", "id": "I1", "implication": "评价一个 AI for Science 项目时，应同时审计其数据可发现性、跨学科接口和社区能力建设，而不是只评估模型 benchmark。", "summary": "Lawrence & Montgomery、Berens 等、AI+MPS 和 AI4OS 从不同入口汇聚到同一判断：前者把科学采用类比为技术扩散，指出 COVID-19 中英国 AI 几乎未进入实际响应；Berens 等将 AI for Science 定义为需要研究者的交汇点；AI+MPS 把社区、数据与教育列为三大支柱；AI4OS 要求把数据上下文和先验知识作为团队间传递的产证。差异在于制度落点：有的是国家层面资金与培训，有的是开放科学伦理；共同点是均未提供定量效果检验。", "title": "算法突破之外，跨学科社区、数据治理和开放机制构成 AI for Science 落地的前置条件；只投模型开发难以跨越科学采用鸿沟。" }, { "boundary": "结论主要来自材料科学和生物/化学领域综述，未证明对所有自然科学成立；“充分性”目前没有可操作定量阈值，且 A-Lab 新颖性本身存在争议。", "claimId": "C2", "confidence": "supported", "evidencePaperIds": [ "openalex:W4402901051", "openalex:W4406828533" ], "explanation": "Lu 等以 AlphaFold 的约 10 万条序列-结构对和 A-Lab 的自动化合成说明大规模数据能带来质变；Sci-LLM 综述则指出预训练语料远小于通用 LLM，且多模态对齐数据稀缺。两者差异在于材料领域强调样本、数据集、利益相关者三层治理，生物化学领域强调多尺度和跨模态数据缺口；共同点是均把数据准备视为研究议程而非附属工程。", "id": "I2", "implication": "资源应优先投向标准化、充分性评估和社区数据共享，而不是在数据薄弱的任务上无限制扩大模型参数。", "summary": "Lu 等以 AlphaFold 的约 10 万条序列-结构对和 A-Lab 的自动化合成说明大规模数据能带来质变；Sci-LLM 综述则指出预训练语料远小于通用 LLM，且多模态对齐数据稀缺。两者差异在于材料领域强调样本、数据集、利益相关者三层治理，生物化学领域强调多尺度和跨模态数据缺口；共同点是均把数据准备视为研究议程而非附属工程。", "title": "在材料与生物化学场景中，FAIR 原则是必要但不充分的数据底线；AI-ready 还要求针对任务检验数据充分性，数据规模与质量常先于模型架构成为性能瓶颈。" }, { "boundary": "ARU 标准尚未在大规模实践中检验；AlphaFold 案例来自 2021 年版本，部分限制可能随版本改善；不适用于无法提供稳定可解释视角的纯黑箱模型。", "claimId": "C3", "confidence": "supported", "evidencePaperIds": [ "arxiv:2406.10557", "openalex:W3205208140", "arxiv:2509.02661" ], "explanation": "Mengaldo 提出 XAI for Science 框架，要求领域专家把机器视角与自身知识比较，在分歧处生成可检验解释；Perrakis/Sixma 通过 AlphaFold 的 CENP-E、Mad2 等具体误判说明高置信度不等于生物学真实；AI+MPS 把可解释性和科学反哺 AI 并列为制度方向。三者对可解释性作用的强调一致，但对人类介入的必要程度不同：Mengaldo 保留人类复杂推理优先，结构生物学案例把实验作为最终仲裁，AI+MPS 更关注方法可审计性。", "id": "I3", "implication": "科学论文中使用 AI 辅助发现时，应报告解释质量、再现实效和实验验证链，而不是仅展示预测分数；评审人应要求 ARU 类门槛或等价的实验证据。", "summary": "Mengaldo 提出 XAI for Science 框架，要求领域专家把机器视角与自身知识比较，在分歧处生成可检验解释；Perrakis/Sixma 通过 AlphaFold 的 CENP-E、Mad2 等具体误判说明高置信度不等于生物学真实；AI+MPS 把可解释性和科学反哺 AI 并列为制度方向。三者对可解释性作用的强调一致，但对人类介入的必要程度不同：Mengaldo 保留人类复杂推理优先，结构生物学案例把实验作为最终仲裁，AI+MPS 更关注方法可审计性。", "title": "仅靠事后解释不足以把 AI 输出变成科学知识；可解释性必须与准确性、可复现性和可理解性共同构成门槛，并经实验或专家分歧裁决。" }, { "boundary": "案例多为作者自选或系统自报，缺少跨系统的统一失败率和成本统计；闭环证据集中在数字策略或特定生物/化学任务，尚未推广到大规模物理实验。", "claimId": "C4", "confidence": "supported", "evidencePaperIds": [ "arxiv:2508.14111", "arxiv:2604.27297", "arxiv:2608.02775", "arxiv:2607.09025" ], "explanation": "Agentic Science 综述以 The Virtual Lab、OriGene 等案例说明特定任务上已能闭环；MCI 在 10 个符号回归基准上降低误差并改善外推；BLAZE 展示了从假设到论文的证据链管理；进化智能框架强调保留失败候选。共同趋势是把科学工作流作为系统对象，但差异明显：MCI 只在数字基准上验证，BLAZE 明确止于 D0 数字阶段，Agentic 综述承认缺少失败率和复现统计，EI 路线图未给出运行系统。", "id": "I4", "implication": "不应把代理科学的宣传当作普适能力，投资和评估应区分数字阶段验证、受控实验和真实实验室闭环三个层级。", "summary": "Agentic Science 综述以 The Virtual Lab、OriGene 等案例说明特定任务上已能闭环；MCI 在 10 个符号回归基准上降低误差并改善外推；BLAZE 展示了从假设到论文的证据链管理；进化智能框架强调保留失败候选。共同趋势是把科学工作流作为系统对象，但差异明显：MCI 只在数字基准上验证，BLAZE 明确止于 D0 数字阶段，Agentic 综述承认缺少失败率和复现统计，EI 路线图未给出运行系统。", "title": "AI for Science 正从单点预测模型转向多代理自主发现和闭环实验，但当前成功主要是数字环境或受控基准中的存在性证明，尚不能声称物理实验自动化已经成熟。" }, { "boundary": "性能数字多转引自原始文献，未统一复现；正面案例有选择偏差；高维非线性问题和缺少误差保证限制其部署可靠性。", "claimId": "C5", "confidence": "single_source", "evidencePaperIds": [ "openalex:W4412642369" ], "explanation": "这是单一 PINN 综述的整理：作者将架构、自适应采样、损失设计与特征嵌入组件化，并列出 DeepXDE 等五个框架；应用覆盖流体、固体、电磁和光学。它没有做过跨方法定量比较，也没有证明 PINN 可替代传统求解器。", "id": "I5", "implication": "科研工作者可将 PINN 视为在已知物理约束下减少数据需求的候选工具，但在关键工程决策前应进行与传统数值方法的受控对比。", "summary": "这是单一 PINN 综述的整理：作者将架构、自适应采样、损失设计与特征嵌入组件化，并列出 DeepXDE 等五个框架；应用覆盖流体、固体、电磁和光学。它没有做过跨方法定量比较，也没有证明 PINN 可替代传统求解器。", "title": "PINNs 通过把 PDE 残差嵌入损失函数，在逆问题、数据稀缺和复杂几何等场景表现出作为数值求解器补充的潜力，但高维、可扩展性和可靠性仍是未解挑战。" }, { "boundary": "仅覆盖生物/化学 Transformer 序列模型，排除图神经网络、扩散模型和数学语言；内容截止 2024 年中，模型更新可能改变部分结论。", "claimId": "C6", "confidence": "single_source", "evidencePaperIds": [ "openalex:W4406828533" ], "explanation": "该综述把科学语言统一为序列建模框架，指出 ProGen 等模型规模远小于通用 LLM，且生成任务的 validity/diversity 等指标只是湿实验的间接信号。这解释了为什么 Sci-LLM 的论文演示多、可验证工具少。", "id": "I6", "implication": "不要只凭公开基准分数选择科学 LLM；在选型和发表时，应明确报告是否经过湿实验或与实验数据库的独立验证。", "summary": "该综述把科学语言统一为序列建模框架，指出 ProGen 等模型规模远小于通用 LLM，且生成任务的 validity/diversity 等指标只是湿实验的间接信号。这解释了为什么 Sci-LLM 的论文演示多、可验证工具少。", "title": "生物与化学科学大语言模型已覆盖文本、分子、蛋白质、基因组和多模态五类，但缺少大学后水平基准，计算指标不能替代湿实验有效性。" }, { "boundary": "案例来自 2021 年版本，未经系统抽样，无法量化错误率；后续版本可能改善多聚体和配体处理，但“缺少能量与动态状态、需要实验仲裁”的逻辑边界可能仍然存在。", "claimId": "C7", "confidence": "single_source", "evidencePaperIds": [ "openalex:W3205208140" ], "explanation": "Perrakis 和 Sixma 用共进化与注意力机制解释 AlphaFold 的成功，同时用配体缺失、多聚体界面和构象状态问题展示其边界；血红蛋白案例说明模型学到的是统计接触而非能量最小化。该单一来源的证据直接支持“高置信度不等于正确”。", "id": "I7", "implication": "使用 AlphaFold 时，应把它作为假设生成器，实验设计应在预测结构之外增加构象验证或功能实验；资助机构不宜削减实验结构生物学投入。", "summary": "Perrakis 和 Sixma 用共进化与注意力机制解释 AlphaFold 的成功，同时用配体缺失、多聚体界面和构象状态问题展示其边界；血红蛋白案例说明模型学到的是统计接触而非能量最小化。该单一来源的证据直接支持“高置信度不等于正确”。", "title": "AlphaFold2 预测结构并不天然代表生物学真实构象；CENP-E 和 Mad2 等预测与实验不符，说明实验结构生物学仍负责判定功能相关的状态。" }, { "boundary": "实验仅覆盖 1000 样本、加性同方差噪声和静态因果图；非线性仅限单调和周期函数；PCMCI 固定 Pearson 检验，无法区分方法缺陷与检验选择。", "claimId": "C8", "confidence": "single_source", "evidencePaperIds": [ "arxiv:2104.08043" ], "explanation": "该基准用可参数化的生成过程合成 200 个 SCM，观察到潜变量使 F1 和 SHD 方向不一致，非线性对线性方法打击最大，瞬时效应导致 Granger 漏报滞后 0 边、PCMCI 产生伴随假阳性。这提醒使用者不能默认某方法在所有时间序列上稳健。", "id": "I8", "implication": "在应用时间序列因果发现前，应先用类似生成框架检查假设符合度并扫描超参数；结果解释必须同时报告 F1 与 SHD，不能仅凭单一指标。", "summary": "该基准用可参数化的生成过程合成 200 个 SCM，观察到潜变量使 F1 和 SHD 方向不一致，非线性对线性方法打击最大，瞬时效应导致 Granger 漏报滞后 0 边、PCMCI 产生伴随假阳性。这提醒使用者不能默认某方法在所有时间序列上稳健。", "title": "时间序列因果发现方法的性能对因果充分性、线性和无瞬时效应等假设高度敏感，且超参数选择（如 PCMCI 的 p 值阈值）能造成与假设违反相当的 F1 变化。" } ], "language": "zh-CN", "narrativeSections": [ { "blocks": [ { "claimIds": [ "C9", "C10", "C11", "C12" ], "epistemicStatus": "cross_paper_synthesis", "evidencePaperIds": [ "arxiv:2505.03977", "arxiv:2501.11847", "arxiv:2401.11839", "arxiv:2607.09025" ], "id": "S1-B1", "role": "answer", "text": "现有研究主要从四类问题入口展开：一是评估与基准问题，SRBench 更新显示没有任何单一符号回归算法在所有数据集上全面领先，且高复杂度模型可能只是拟合噪声，能耗与超参数搜索空间对排名有实质影响；二是训练基础设施问题，科学 Transformer 模型受 AI 内存墙制约，现有通用内存优化技术未充分应用于科学架构，需要针对 Evoformer、SE(3)-Transformer 等定制优化；三是领域扩展与效度问题，社会科学中 LLM 作为工具在速度、成本、可重复性上占优，但在效度、数值分析和机制解释方面较弱，尚不能替代社会科学家；四是发现过程的累积机制问题，演化智能主张累积科学发现的关键不仅在于候选解优化，还在于保留搜索历史和失败候选作为科学证据。" }, { "claimIds": [ "C9", "C10" ], "epistemicStatus": "cross_paper_synthesis", "evidencePaperIds": [ "arxiv:2505.03977", "arxiv:2501.11847" ], "id": "S1-B2", "role": "comparison", "text": "对比两个分支可见问题入口的差异：基准分支以“哪个方法更好”为问题，却得出没有单一算法全面领先的结论，并指出某些高复杂度模型的高 R² 可能只是拟合噪声；内存分支则以“模型能否跑起来”为问题，指出科学 Transformer 的训练受参数增长超过加速器内存容量限制，通用内存优化技术未充分定制，AlphaFold2 的序列维动态并行是针对性优化案例。前者在算法层面揭示性能评价的场景依赖性，后者在系统层面指出扩展瓶颈，二者分别对应评价与训练两个不同阶段，共同说明研究版图不是单一算法竞赛，而是由评测、计算、数据等多重基础设施共同决定。" }, { "claimIds": [ "C11", "C12" ], "epistemicStatus": "paper_finding", "evidencePaperIds": [ "arxiv:2401.11839", "arxiv:2607.09025" ], "id": "S1-B3", "role": "evidence", "text": "在社会科学分支，arxiv:2401.11839 的综述明确列出 LLM 相比传统方法在速度、成本、不可重复性、可扩展性等维度更优，但在效度和数值分析上更差；演化智能框架（arxiv:2607.09025）则从逻辑上论证失败候选和谱系记录可以转化为科学证据，但没有提供实证系统。这些证据形态本身反映了版图的不平衡：基准和工具类工作能提供可量化结果，而领域扩展和累积机制仍处于框架或警示阶段。" }, { "claimIds": [ "C9", "C10", "C11", "C12" ], "epistemicStatus": "editorial_inference", "evidencePaperIds": [ "arxiv:2505.03977", "arxiv:2501.11847", "arxiv:2401.11839", "arxiv:2607.09025" ], "id": "S1-B4", "role": "implication", "text": "理解研究版图后，读者不应把 AI for Science 看成单一方法竞争，而应看到四类问题之间的依赖：评估基准影响方法选择，训练基础设施限制模型规模，领域扩展要求重新定义效度，累积机制决定长期科研效率。未来工作若要推进，需要在至少一个入口给出可操作的验证标准。" } ], "confidence": "mixed", "evidencePaperIds": [ "arxiv:2505.03977", "arxiv:2501.11847", "arxiv:2401.11839", "arxiv:2607.09025" ], "id": "S1", "kind": "research_landscape", "moduleId": "M3", "readerQuestion": "现有研究主要从哪些问题入口展开？", "title": "当前研究版图" }, { "blocks": [ { "claimIds": [ "C2", "C5", "C7", "C8", "C9" ], "epistemicStatus": "editorial_inference", "evidencePaperIds": [ "openalex:W4402901051", "openalex:W4412642369", "openalex:W3205208140", "arxiv:2104.08043", "arxiv:2505.03977" ], "id": "S2-B1", "role": "answer", "text": "当前证据没有回答：数据充分性所需的最小规模与质量阈值；PINNs 在高维、非线性和可靠性上的失败率与可扩展边界；AlphaFold2 高置信预测的错误比例及随版本更新的变化；时间序列因果发现方法在更大样本、异方差或时变图下的表现；符号回归能耗的实测值及算法差异的统计显著性。这些空白使多数结论只能作为方向性判断，而非可操作规范。" }, { "claimIds": [ "C2", "C6" ], "epistemicStatus": "cross_paper_synthesis", "evidencePaperIds": [ "openalex:W4402901051", "openalex:W4406828533" ], "id": "S2-B2", "role": "comparison", "text": "对比材料科学与生物/化学两条线索：openalex:W4402901051 强调材料数据不足和 FAIR 不够，但未给出充分性定量标准；openalex:W4406828533 强调科学 LLM 评估滞后，但仅覆盖 Transformer 且内容已部分过时。两者共同指向支撑层缺口，却分别受限于领域和模型范围，不能合并为跨领域结论。" }, { "claimIds": [ "C7", "C8" ], "epistemicStatus": "paper_finding", "evidencePaperIds": [ "openalex:W3205208140", "arxiv:2104.08043" ], "id": "S2-B3", "role": "evidence", "text": "openalex:W3205208140 记录了两个具体不符案例：CENP-E 的卷曲螺旋被预测为球状蛋白，而电镜显示为延伸结构；Mad2 被预测为结合态而非游离态。这是 AlphaFold2 高置信预测可能生物学错误的直接证据。此外，arxiv:2104.08043 的基准只使用了 1000 样本、加性同方差噪声和静态因果图，因此不能外推到所有真实时间序列场景。" }, { "claimIds": [ "C1", "C2", "C3", "C4", "C5", "C6", "C7", "C8", "C9", "C10", "C11", "C12" ], "epistemicStatus": "editorial_inference", "evidencePaperIds": [ "openalex:W4401726605", "openalex:W4412642369", "openalex:W4406828533", "arxiv:2508.14111", "arxiv:2509.02661", "openalex:W3205208140", "openalex:W4323697696", "arxiv:2604.27297", "arxiv:2607.09025", "arxiv:2310.18852", "arxiv:2406.10557", "arxiv:2401.11839", "arxiv:2501.11847", "arxiv:2608.02775", "openalex:W4402901051", "arxiv:2505.03977", "arxiv:2104.08043" ], "id": "S2-B4", "role": "boundary", "text": "需要区分两类未知：一类是尚未测量但原则上可测量（例如更大规模的基准、更全面的能耗统计）；另一类是当前方法可能本质上无法回答（例如缺乏统一因果模型时，任何因果发现方法的性能上限）。前一类可通过后续实验解决，后一类需要新的理论框架。现有证据不足以支持任何“AI 已普遍取代人类科学家”的强主张。" } ], "confidence": "mixed", "evidencePaperIds": [ "openalex:W4402901051", "openalex:W4412642369", "openalex:W3205208140", "arxiv:2104.08043", "arxiv:2505.03977", "openalex:W4406828533", "openalex:W4401726605", "arxiv:2508.14111", "arxiv:2509.02661", "openalex:W4323697696", "arxiv:2604.27297", "arxiv:2607.09025", "arxiv:2310.18852", "arxiv:2406.10557", "arxiv:2401.11839", "arxiv:2501.11847", "arxiv:2608.02775" ], "id": "S2", "kind": "evidence_boundaries", "moduleId": "M4", "readerQuestion": "当前证据没有回答什么？", "title": "这些结论能相信到什么程度" } ], "papers": [ { "citation": { "capturedAt": "2026-08-13", "count": 16, "source": "openalex" }, "firstPublished": "2024-08-01", "id": "openalex:W4401726605", "identifiers": { "doi": "10.1098/rsos.231130", "openalex": "W4401726605" }, "importance": 2, "importanceReason": "Addresses the diffusion of AI across sciences and defines an open data science framework with supply chains of ideas, technology transfer, and data stewardship.", "noteHref": "papers/d254385381c88895cadb868e01def6ae7130f88bc615fa64f4591776d399c875.md", "notePath": "papers/d254385381c88895cadb868e01def6ae7130f88bc615fa64f4591776d399c875.md", "pdfHref": "papers/d254385381c88895cadb868e.pdf", "role": "foundation", "sourceUrl": "https://doi.org/10.1098/rsos.231130", "tags": [ "理论与立场" ], "title": "Accelerating AI for science: open data science for science", "venue": "openalex" }, { "citation": { "capturedAt": "2026-08-13", "count": 158, "source": "openalex" }, "firstPublished": "2025-07-24", "id": "openalex:W4412642369", "identifiers": { "doi": "10.1007/s10462-025-11322-7", "openalex": "W4412642369" }, "importance": 2, "importanceReason": "Comprehensive review of physics-informed neural networks for solving PDEs, integrating physical laws into deep learning.", "noteHref": "papers/c07e03bcfeb8bf1c30f5de544a4c359f109e85572840ea2ae22d0836948d7247.md", "notePath": "papers/c07e03bcfeb8bf1c30f5de544a4c359f109e85572840ea2ae22d0836948d7247.md", "pdfHref": "papers/c07e03bcfeb8bf1c30f5de54.pdf", "role": "application", "sourceUrl": "https://doi.org/10.1007/s10462-025-11322-7", "tags": [ "综述与文献回顾" ], "title": "Physics-informed neural networks for PDE problems: a comprehensive review", "venue": "openalex" }, { "citation": { "capturedAt": "2026-08-13", "count": 82, "source": "openalex" }, "firstPublished": "2024-01-26", "id": "openalex:W4406828533", "identifiers": { "arxiv": "2401.14656", "doi": "10.1145/3715318", "openalex": "W4406828533" }, "importance": 2, "importanceReason": "Survey of scientific large language models in biology and chemistry, covering architectures, datasets, and evaluation.", "noteHref": "papers/86e50184151d786fa6483233c55718d1697fa7ab68b53bbee2daa651a50366bd.md", "notePath": "papers/86e50184151d786fa6483233c55718d1697fa7ab68b53bbee2daa651a50366bd.md", "pdfHref": "papers/86e50184151d786fa6483233.pdf", "role": "application", "sourceUrl": "https://doi.org/10.1145/3715318", "tags": [ "综述与文献回顾" ], "title": "Scientific Large Language Models: A Survey on Biological & Chemical Domains", "venue": "openalex" }, { "citation": { "capturedAt": "2026-08-13", "source": "arxiv", "unavailableReason": "可信来源未提供引用数" }, "firstPublished": "2026-04-30", "id": "arxiv:2604.27297", "identifiers": { "arxiv": "2604.27297" }, "importance": 2, "importanceReason": "Presents machine collective intelligence for autonomous discovery of governing equations, reducing extrapolation error by orders of magnitude.", "noteHref": "papers/1d1fec14a2111e177b096a1e230d9c34bafbb7b1c378e145c17a25e2f1f56f75.md", "notePath": "papers/1d1fec14a2111e177b096a1e230d9c34bafbb7b1c378e145c17a25e2f1f56f75.md", "pdfHref": "papers/1d1fec14a2111e177b096a1e.pdf", "role": "foundation", "sourceUrl": "https://arxiv.org/abs/2604.27297v1", "tags": [ "cs.AI" ], "title": "Machine Collective Intelligence for Explainable Scientific Discovery", "venue": "arxiv" }, { "citation": { "capturedAt": "2026-08-13", "count": 5, "source": "arxiv" }, "firstPublished": "2025-08-18", "id": "arxiv:2508.14111", "identifiers": { "arxiv": "2508.14111", "doi": "10.48550/arxiv.2508.14111", "openalex": "W4415238177" }, "importance": 2, "importanceReason": "Survey on agentic science, positioning AI as autonomous research partners across domains, with a framework for autonomous discovery.", "noteHref": "papers/2010a5634d5d2583a9d0ca57ae92118191fe465cce40c5f0b6575cbcf12560f5.md", "notePath": "papers/2010a5634d5d2583a9d0ca57ae92118191fe465cce40c5f0b6575cbcf12560f5.md", "pdfHref": "papers/2010a5634d5d2583a9d0ca57.pdf", "role": "foundation", "sourceUrl": "https://arxiv.org/abs/2508.14111v2", "tags": [ "cs.LG" ], "title": "From AI for Science to Agentic Science: A Survey on Autonomous Scientific Discovery", "venue": "arxiv" }, { "citation": { "capturedAt": "2026-08-13", "source": "arxiv", "unavailableReason": "可信来源未提供引用数" }, "firstPublished": "2025-09-02", "id": "arxiv:2509.02661", "identifiers": { "arxiv": "2509.02661" }, "importance": 2, "importanceReason": "NSF workshop report on the future of AI and mathematical/physical sciences, outlining strategic priorities for research and education.", "noteHref": "papers/25390665536831ec02277bcf30bf3ec60b7dffad10e5555e83dfa0571d113f27.md", "notePath": "papers/25390665536831ec02277bcf30bf3ec60b7dffad10e5555e83dfa0571d113f27.md", "pdfHref": "papers/25390665536831ec02277bcf.pdf", "role": "application", "sourceUrl": "https://arxiv.org/abs/2509.02661v3", "tags": [ "cs.AI" ], "title": "The Future of Artificial Intelligence and the Mathematical and Physical Sciences (AI+MPS)", "venue": "arxiv" }, { "citation": { "capturedAt": "2026-08-13", "count": 202, "source": "openalex" }, "firstPublished": "2021-10-20", "id": "openalex:W3205208140", "identifiers": { "doi": "10.15252/embr.202154046", "openalex": "W3205208140" }, "importance": 2, "importanceReason": "Highlights AlphaFold as a groundbreaking AI application in biology, with cautionary notes on its limitations.", "noteHref": "papers/e4bbeb6b5593b10e88f991ce10a7c489af813f1cfd42ac00e9c074dddfa52444.md", "notePath": "papers/e4bbeb6b5593b10e88f991ce10a7c489af813f1cfd42ac00e9c074dddfa52444.md", "pdfHref": "papers/e4bbeb6b5593b10e88f991ce.pdf", "role": "foundation", "sourceUrl": "https://doi.org/10.15252/embr.202154046", "tags": [ "理论与立场" ], "title": "AI revolutions in biology", "venue": "openalex" }, { "citation": { "capturedAt": "2026-08-13", "count": 9, "source": "openalex" }, "firstPublished": "2023-03-07", "id": "openalex:W4323697696", "identifiers": { "arxiv": "2303.04217", "doi": "10.48550/arxiv.2303.04217", "openalex": "W4323697696" }, "importance": 2, "importanceReason": "Dagstuhl seminar report on AI for science, bridging data-driven and mechanistic modelling to address complex systems.", "noteHref": "papers/8c314053baba7f78e916ebbd09d18ac08fa3e9166c777589dec46f96c1749bdb.md", "notePath": "papers/8c314053baba7f78e916ebbd09d18ac08fa3e9166c777589dec46f96c1749bdb.md", "pdfHref": "papers/8c314053baba7f78e916ebbd.pdf", "role": "foundation", "sourceUrl": "http://arxiv.org/abs/2303.04217", "tags": [ "理论与立场" ], "title": "AI for Science: An Emerging Agenda", "venue": "openalex" }, { "citation": { "capturedAt": "2026-08-13", "source": "arxiv", "unavailableReason": "可信来源未提供引用数" }, "firstPublished": "2026-07-10", "id": "arxiv:2607.09025", "identifiers": { "arxiv": "2607.09025" }, "importance": 2, "importanceReason": "Review of evolutionary intelligence for cumulative scientific discovery, proposing a five-dimensional framework.", "noteHref": "papers/2b782e7e81cf876ca5c8789ed34a735f4908ada9ccbddfeceebf081dbecdd287.md", "notePath": "papers/2b782e7e81cf876ca5c8789ed34a735f4908ada9ccbddfeceebf081dbecdd287.md", "pdfHref": "papers/2b782e7e81cf876ca5c8789e.pdf", "role": "foundation", "sourceUrl": "https://arxiv.org/abs/2607.09025v1", "tags": [ "cs.NE" ], "title": "Evolutionary Intelligence for Scientific Discovery: From Evolutionary Computation to Cumulative Discovery Systems", "venue": "arxiv" }, { "citation": { "capturedAt": "2026-08-13", "source": "arxiv", "unavailableReason": "可信来源未提供引用数" }, "firstPublished": "2023-10-28", "id": "arxiv:2310.18852", "identifiers": { "arxiv": "2310.18852" }, "importance": 2, "importanceReason": "Introduces AI for Open Science as a multi-agent extension, ensuring broad knowledge translation and ethical sharing.", "noteHref": "papers/0244e2f995b679ad3035a03cf002bcc8bfe7cf7e87014045b8dc7f819854bf7b.md", "notePath": "papers/0244e2f995b679ad3035a03cf002bcc8bfe7cf7e87014045b8dc7f819854bf7b.md", "pdfHref": "papers/0244e2f995b679ad3035a03c.pdf", "role": "foundation", "sourceUrl": "https://arxiv.org/abs/2310.18852v2", "tags": [ "cs.AI" ], "title": "AI for Open Science: A Multi-Agent Perspective for Ethically Translating Data to Knowledge", "venue": "arxiv" }, { "citation": { "capturedAt": "2026-08-13", "source": "arxiv", "unavailableReason": "可信来源未提供引用数" }, "firstPublished": "2024-06-15", "id": "arxiv:2406.10557", "identifiers": { "arxiv": "2406.10557" }, "importance": 2, "importanceReason": "Argues for explainable AI in science, where interpretability guides scientific hypotheses and new knowledge.", "noteHref": "papers/33450a7130902c6f38cc226f647f561123c98a88e677a67d6876b20040811052.md", "notePath": "papers/33450a7130902c6f38cc226f647f561123c98a88e677a67d6876b20040811052.md", "pdfHref": "papers/33450a7130902c6f38cc226f.pdf", "role": "foundation", "sourceUrl": "https://arxiv.org/abs/2406.10557v5", "tags": [ "cs.AI" ], "title": "Explain the Black Box for the Sake of Science: the Scientific Method in the Era of Generative Artificial Intelligence", "venue": "arxiv" }, { "citation": { "capturedAt": "2026-08-13", "source": "arxiv", "unavailableReason": "可信来源未提供引用数" }, "firstPublished": "2024-01-22", "id": "arxiv:2401.11839", "identifiers": { "arxiv": "2401.11839" }, "importance": 2, "importanceReason": "Surveys AI for social science and social science of AI, providing a unified framework for studying AI's societal role.", "noteHref": "papers/1ec8328b51ab7411e7e247b969eec6d541db340b71f3d127a4bb06869c89bf98.md", "notePath": "papers/1ec8328b51ab7411e7e247b969eec6d541db340b71f3d127a4bb06869c89bf98.md", "pdfHref": "papers/1ec8328b51ab7411e7e247b9.pdf", "role": "application", "sourceUrl": "https://arxiv.org/abs/2401.11839v1", "tags": [ "cs.CL" ], "title": "AI for social science and social science of AI: A Survey", "venue": "arxiv" }, { "citation": { "capturedAt": "2026-08-13", "source": "arxiv", "unavailableReason": "可信来源未提供引用数" }, "firstPublished": "2025-01-21", "id": "arxiv:2501.11847", "identifiers": { "arxiv": "2501.11847" }, "importance": 3, "importanceReason": "作为核心精读论文，为研究问题提供直接证据。", "noteHref": "papers/54d839ae737205696bb3d17a57b2be8b1df4d4cc336787eb5ad5311df6ac6f67.md", "notePath": "papers/54d839ae737205696bb3d17a57b2be8b1df4d4cc336787eb5ad5311df6ac6f67.md", "pdfHref": "papers/54d839ae737205696bb3d17a.pdf", "role": "综述与文献回顾", "sourceUrl": "https://arxiv.org/abs/2501.11847v2", "tags": [ "cs.LG" ], "title": "A Survey on Memory-Efficient Transformer-Based Model Training in AI for Science", "venue": "arxiv" }, { "citation": { "capturedAt": "2026-08-13", "source": "arxiv", "unavailableReason": "可信来源未提供引用数" }, "firstPublished": "2026-08-03", "id": "arxiv:2608.02775", "identifiers": { "arxiv": "2608.02775" }, "importance": 3, "importanceReason": "作为核心精读论文，为研究问题提供直接证据。", "noteHref": "papers/adecf25c991854b3e092b23a6b418e91091aede39bf6eef0fa2598c75d805d27.md", "notePath": "papers/adecf25c991854b3e092b23a6b418e91091aede39bf6eef0fa2598c75d805d27.md", "pdfHref": "papers/adecf25c991854b3e092b23a.pdf", "role": "方法与系统", "sourceUrl": "https://arxiv.org/abs/2608.02775v1", "tags": [ "cs.AI" ], "title": "Towards a new paradigm of scientific discovery with socialized artificial intelligence", "venue": "arxiv" }, { "citation": { "capturedAt": "2026-08-13", "count": 16, "source": "openalex" }, "firstPublished": "2024-09-27", "id": "openalex:W4402901051", "identifiers": { "doi": "10.1038/s41597-024-03821-z", "openalex": "W4402901051" }, "importance": 3, "importanceReason": "作为核心精读论文，为研究问题提供直接证据。", "noteHref": "papers/33422cf366c2fecc34ca5a7dbec40c0de3bd059a8101ac25ad23bdd0d87da1b5.md", "notePath": "papers/33422cf366c2fecc34ca5a7dbec40c0de3bd059a8101ac25ad23bdd0d87da1b5.md", "pdfHref": "papers/33422cf366c2fecc34ca5a7d.pdf", "role": "理论与立场", "sourceUrl": "https://doi.org/10.1038/s41597-024-03821-z", "tags": [ "理论与立场" ], "title": "Unleashing the power of AI in science-key considerations for materials data preparation", "venue": "openalex" }, { "citation": { "capturedAt": "2026-08-13", "source": "arxiv", "unavailableReason": "可信来源未提供引用数" }, "firstPublished": "2021-10-05", "id": "arxiv:2110.01831", "identifiers": { "arxiv": "2110.01831" }, "importance": 3, "importanceReason": "作为核心精读论文，为研究问题提供直接证据。", "noteHref": "papers/fae5116b7f2c337341c5d0c4ee3d28cbab64ea62842cdaa629c27ca6b45ed72d.md", "notePath": "papers/fae5116b7f2c337341c5d0c4ee3d28cbab64ea62842cdaa629c27ca6b45ed72d.md", "pdfHref": "papers/fae5116b7f2c337341c5d0c4.pdf", "role": "理论与立场", "sourceUrl": "https://arxiv.org/abs/2110.01831v1", "tags": [ "cs.AI" ], "title": "The Artificial Scientist: Logicist, Emergentist, and Universalist Approaches to Artificial General Intelligence", "venue": "arxiv" }, { "citation": { "capturedAt": "2026-08-13", "source": "arxiv", "unavailableReason": "可信来源未提供引用数" }, "firstPublished": "2025-05-06", "id": "arxiv:2505.03977", "identifiers": { "arxiv": "2505.03977" }, "importance": 2, "importanceReason": "concept_coverage=0.331; evidence_availability=1.000; lexical_relevance=0.246; recency=1.000; screening_relevance=1.000; semantic_relevance=0.423; source_consistency=0.500", "noteHref": "papers/5a9ff68991d29dc8e7fd9f97d908d011c949f58fa6d29cdaaacaad2d4fa4460a.md", "notePath": "papers/5a9ff68991d29dc8e7fd9f97d908d011c949f58fa6d29cdaaacaad2d4fa4460a.md", "pdfHref": "papers/5a9ff68991d29dc8e7fd9f97.pdf", "role": "supplementary", "sourceUrl": "https://arxiv.org/abs/2505.03977v1", "tags": [ "cs.LG" ], "title": "Call for Action: towards the next generation of symbolic regression benchmark", "venue": "arxiv" }, { "citation": { "capturedAt": "2026-08-13", "source": "arxiv", "unavailableReason": "可信来源未提供引用数" }, "firstPublished": "2026-03-30", "id": "arxiv:2603.28361", "identifiers": { "arxiv": "2603.28361" }, "importance": 2, "importanceReason": "concept_coverage=0.451; evidence_availability=1.000; lexical_relevance=0.012; recency=1.000; screening_relevance=0.800; semantic_relevance=0.556; source_consistency=0.500", "noteHref": "papers/6e5cd35f3468a0bccffd84fef0ba4a80b52813f0be22b96088237e669beaa970.md", "notePath": "papers/6e5cd35f3468a0bccffd84fef0ba4a80b52813f0be22b96088237e669beaa970.md", "pdfHref": "papers/6e5cd35f3468a0bccffd84fe.pdf", "role": "supplementary", "sourceUrl": "https://arxiv.org/abs/2603.28361v1", "tags": [ "cs.AI" ], "title": "Deep Research of Deep Research: From Transformer to Agent, From AI to AI for Science", "venue": "arxiv" }, { "citation": { "capturedAt": "2026-08-13", "source": "arxiv", "unavailableReason": "可信来源未提供引用数" }, "firstPublished": "2021-04-16", "id": "arxiv:2104.08043", "identifiers": { "arxiv": "2104.08043" }, "importance": 2, "importanceReason": "concept_coverage=0.306; evidence_availability=1.000; lexical_relevance=0.246; recency=0.667; screening_relevance=0.700; semantic_relevance=0.374; source_consistency=0.500", "noteHref": "papers/1ce5eb1743bba16617ab95dadcd1caff70c2e87802e4e6f72ec249466fd4c7cd.md", "notePath": "papers/1ce5eb1743bba16617ab95dadcd1caff70c2e87802e4e6f72ec249466fd4c7cd.md", "pdfHref": "papers/1ce5eb1743bba16617ab95da.pdf", "role": "supplementary", "sourceUrl": "https://arxiv.org/abs/2104.08043v1", "tags": [ "stat.ML" ], "title": "Data Generating Process to Evaluate Causal Discovery Techniques for Time Series Data", "venue": "arxiv" } ], "privacy": { "defaultPrivate": true, "forbiddenTerms": [] }, "repositories": [ { "abstract": "Scientific machine learning (SciML) benchmarks, AI for science, and (differential) equation solvers. Covers Julia, Python (PyTorch, Jax), MATLAB, R", "alternate_pdf_urls": [], "authors": [], "canonical_id": "github:sciml/scimlbenchmarks.jl", "first_published": null, "identifiers": { "github": "sciml/scimlbenchmarks.jl" }, "institutions": [], "kind": "repository", "metrics": { "forks": 112, "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 0, "graph_foundation": 0.0, "graph_frontier": 0.275, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "landscape_relevance_class": "contextual", "landscape_relevance_reasons": [ "screening_contextual_value" ], "paper_url": "https://docs.sciml.ai/SciMLBenchmarksOutput/stable/", "pushed_at": "2026-08-12T09:58:19Z", "stars": 344, "topic_cluster": 0, "topic_cluster_coherence": 0.5440023663661402, "topic_cluster_relation": "direct", "topic_cluster_title": "science / scientific / intelligence" }, "pdf_url": null, "provenance": [ "github:sciml/scimlbenchmarks.jl" ], "raw_ref": "SciML/SciMLBenchmarks.jl", "references": [], "source": "github", "source_url": "https://github.com/SciML/SciMLBenchmarks.jl", "tags": [ "ai", "ai-for-science", "benchmark", "dae", "differential-equations", "differentialequations", "jax", "julia", "matlab", "nerual-differential-equations", "neural-ode", "ode", "partial-differential-equations", "pde", "python", "pytorch", "scientific-machine-learning", "sciml", "sde" ], "title": "SciML/SciMLBenchmarks.jl" } ], "researchIntelligence": { "citationGraph": { "coverage": { "availablePaperCount": 6, "status": "available", "totalPaperCount": 19 }, "direction": "citing_to_referenced", "edges": [ { "fromPaperId": "openalex:W4401726605", "source": "openalex", "toPaperId": "openalex:W4323697696" }, { "fromPaperId": "openalex:W4401726605", "source": "verified_reference", "toPaperId": "openalex:W4323697696" } ], "methodNote": "仅显示由精确论文标识确认的集合内引用；缺少边不等于论文之间没有关系。", "nodes": [ { "bridge": 0.0, "community": 0, "firstPublished": "2024-08-01", "foundation": 0.0, "frontier": 0.7812745207316192, "importance": 2, "paperId": "openalex:W4401726605", "structuralRole": "frontier", "tags": [ "理论与立场" ], "title": "Accelerating AI for science: open data science for science" }, { "bridge": 0.0, "community": 0, "firstPublished": "2025-07-24", "foundation": 0.0, "frontier": 0.6950000000000001, "importance": 2, "paperId": "openalex:W4412642369", "structuralRole": "frontier", "tags": [ "综述与文献回顾" ], "title": "Physics-informed neural networks for PDE problems: a comprehensive review" }, { "bridge": 0.0, "community": 0, "firstPublished": "2024-01-26", "foundation": 0.0, "frontier": 0.6323569317546065, "importance": 2, "paperId": "openalex:W4406828533", "structuralRole": "frontier", "tags": [ "综述与文献回顾" ], "title": "Scientific Large Language Models: A Survey on Biological & Chemical Domains" }, { "bridge": 0.0, "community": 0, "firstPublished": "2026-04-30", "foundation": 0.0, "frontier": 0.55, "importance": 2, "paperId": "arxiv:2604.27297", "structuralRole": "frontier", "tags": [ "cs.AI" ], "title": "Machine Collective Intelligence for Explainable Scientific Discovery" }, { "bridge": 0.0, "community": 0, "firstPublished": "2025-08-18", "foundation": 0.0, "frontier": 0.5656961267266897, "importance": 2, "paperId": "arxiv:2508.14111", "structuralRole": "frontier", "tags": [ "cs.LG" ], "title": "From AI for Science to Agentic Science: A Survey on Autonomous Scientific Discovery" }, { "bridge": 0.0, "community": 0, "firstPublished": "2025-09-02", "foundation": 0.0, "frontier": 0.49500000000000005, "importance": 2, "paperId": "arxiv:2509.02661", "structuralRole": "frontier", "tags": [ "cs.AI" ], "title": "The Future of Artificial Intelligence and the Mathematical and Physical Sciences (AI+MPS)" }, { "bridge": 0.0, "community": 1, "firstPublished": "2021-10-20", "foundation": 0.0, "frontier": 0.3960352683688942, "importance": 2, "paperId": "openalex:W3205208140", "structuralRole": "frontier", "tags": [ "理论与立场" ], "title": "AI revolutions in biology" }, { "bridge": 0.0, "community": 0, "firstPublished": "2023-03-07", "foundation": 0.2, "frontier": 0.44924163758320745, "importance": 2, "paperId": "openalex:W4323697696", "structuralRole": "frontier", "tags": [ "理论与立场" ], "title": "AI for Science: An Emerging Agenda" }, { "bridge": 0.0, "community": 0, "firstPublished": "2026-07-10", "foundation": 0.0, "frontier": 0.55, "importance": 2, "paperId": "arxiv:2607.09025", "structuralRole": "frontier", "tags": [ "cs.NE" ], "title": "Evolutionary Intelligence for Scientific Discovery: From Evolutionary Computation to Cumulative Discovery Systems" }, { "bridge": 0.0, "community": 0, "firstPublished": "2023-10-28", "foundation": 0.0, "frontier": 0.385, "importance": 2, "paperId": "arxiv:2310.18852", "structuralRole": "frontier", "tags": [ "cs.AI" ], "title": "AI for Open Science: A Multi-Agent Perspective for Ethically Translating Data to Knowledge" }, { "bridge": 0.0, "community": 0, "firstPublished": "2024-06-15", "foundation": 0.0, "frontier": 0.44000000000000006, "importance": 2, "paperId": "arxiv:2406.10557", "structuralRole": "frontier", "tags": [ "cs.AI" ], "title": "Explain the Black Box for the Sake of Science: the Scientific Method in the Era of Generative Artificial Intelligence" }, { "bridge": 0.0, "community": 0, "firstPublished": "2024-01-22", "foundation": 0.0, "frontier": 0.44000000000000006, "importance": 2, "paperId": "arxiv:2401.11839", "structuralRole": "frontier", "tags": [ "cs.CL" ], "title": "AI for social science and social science of AI: A Survey" }, { "bridge": 0.0, "community": 0, "firstPublished": "2025-01-21", "foundation": 0.0, "frontier": 0.49500000000000005, "importance": 3, "paperId": "arxiv:2501.11847", "structuralRole": "frontier", "tags": [ "cs.LG" ], "title": "A Survey on Memory-Efficient Transformer-Based Model Training in AI for Science" }, { "bridge": 0.0, "community": 0, "firstPublished": "2026-08-03", "foundation": 0.0, "frontier": 0.55, "importance": 3, "paperId": "arxiv:2608.02775", "structuralRole": "frontier", "tags": [ "cs.AI" ], "title": "Towards a new paradigm of scientific discovery with socialized artificial intelligence" }, { "bridge": 0.0, "community": 1, "firstPublished": "2024-09-27", "foundation": 0.0, "frontier": 0.5812745207316192, "importance": 3, "paperId": "openalex:W4402901051", "structuralRole": "frontier", "tags": [ "理论与立场" ], "title": "Unleashing the power of AI in science-key considerations for materials data preparation" }, { "bridge": 0.0, "community": 0, "firstPublished": "2021-10-05", "foundation": 0.0, "frontier": 0.275, "importance": 3, "paperId": "arxiv:2110.01831", "structuralRole": "frontier", "tags": [ "cs.AI" ], "title": "The Artificial Scientist: Logicist, Emergentist, and Universalist Approaches to Artificial General Intelligence" }, { "bridge": null, "community": null, "firstPublished": "2025-05-06", "foundation": null, "frontier": null, "importance": 2, "paperId": "arxiv:2505.03977", "structuralRole": "", "tags": [ "cs.LG" ], "title": "Call for Action: towards the next generation of symbolic regression benchmark" }, { "bridge": null, "community": null, "firstPublished": "2026-03-30", "foundation": null, "frontier": null, "importance": 2, "paperId": "arxiv:2603.28361", "structuralRole": "", "tags": [ "cs.AI" ], "title": "Deep Research of Deep Research: From Transformer to Agent, From AI to AI for Science" }, { "bridge": null, "community": null, "firstPublished": "2021-04-16", "foundation": null, "frontier": null, "importance": 2, "paperId": "arxiv:2104.08043", "structuralRole": "", "tags": [ "stat.ML" ], "title": "Data Generating Process to Evaluate Causal Discovery Techniques for Time Series Data" } ] }, "evidenceMatrix": { "claims": [ { "contradictingPaperIds": [], "id": "C1", "maturity": "supported", "statement": "AI for Science 的核心共识是，其成功不仅依赖算法改进，还取决于跨学科社区建设、数据基础设施和开放科学机制；单纯方法突破不足以解决科学采用障碍。", "supportingPaperIds": [ "openalex:W4323697696", "openalex:W4401726605", "arxiv:2509.02661", "arxiv:2310.18852" ] }, { "contradictingPaperIds": [], "id": "C2", "maturity": "supported", "statement": "在材料科学和生物化学等领域，高质量、大规模且标准化的数据是 AI 模型性能的主要限制因素；仅遵循 FAIR 原则不足以保证 AI 就绪性，还需要评估数据充分性。", "supportingPaperIds": [ "openalex:W4402901051", "openalex:W4406828533" ] }, { "contradictingPaperIds": [], "id": "C3", "maturity": "supported", "statement": "多个分支强调可解释性和人类专家介入对科学发现至关重要；仅靠事后解释不足以保证科学结论，需要结合准确性、可再现性和可理解性标准，并用实验进行裁决。", "supportingPaperIds": [ "arxiv:2406.10557", "openalex:W3205208140", "arxiv:2509.02661" ] }, { "contradictingPaperIds": [], "id": "C4", "maturity": "supported", "statement": "AI for Science 正在从单一预测模型向自主发现、闭环实验的‘代理科学’演进，但当前成功案例（如虚拟实验室、机器集体智能）主要是在受控基准或数字阶段的存在性证明，尚未形成可推广到物理实验的成熟能力。", "supportingPaperIds": [ "arxiv:2508.14111", "arxiv:2604.27297", "arxiv:2608.02775", "arxiv:2607.09025" ] }, { "contradictingPaperIds": [], "id": "C5", "maturity": "single_source", "statement": "物理信息神经网络（PINNs）通过将偏微分方程嵌入损失函数，在流体、固体力学等领域展现出作为数值求解器替代或补充的潜力，但在高维问题、可扩展性和可靠性方面仍存在显著未解决挑战。", "supportingPaperIds": [ "openalex:W4412642369" ] }, { "contradictingPaperIds": [], "id": "C6", "maturity": "single_source", "statement": "生物与化学科学大语言模型已发展出文本、分子、蛋白质、基因组和多模态五类，但评估基础设施滞后：缺乏大学后水平基准，且计算指标不能直接代表湿实验有效性。", "supportingPaperIds": [ "openalex:W4406828533" ] }, { "contradictingPaperIds": [], "id": "C7", "maturity": "single_source", "statement": "AlphaFold2 的高置信度预测并不等于生物学上正确的结构；部分预测（如 CENP-E、Mad2）与实验结构不符，说明实验结构生物学仍是判定生物学真实状态和构象的必要手段。", "supportingPaperIds": [ "openalex:W3205208140" ] }, { "contradictingPaperIds": [], "id": "C8", "maturity": "single_source", "statement": "因果发现基准实验表明，时间序列因果推断方法（如 Granger、PCMCI、DYNOTEARS）的性能对因果充分性、线性、无瞬时效应等假设高度敏感；违反假设时 F1 显著下降，且超参数选择（如 PCMCI 的 p 值阈值）对结果影响强烈。", "supportingPaperIds": [ "arxiv:2104.08043" ] }, { "contradictingPaperIds": [], "id": "C9", "maturity": "single_source", "statement": "更新版符号回归基准（SRBench）显示，没有任何单一符号回归算法在所有数据集上全面领先；能耗、超参数搜索空间和数据噪声对算法排名有实质性影响，且高复杂度模型可能只是拟合噪声。", "supportingPaperIds": [ "arxiv:2505.03977" ] }, { "contradictingPaperIds": [], "id": "C10", "maturity": "single_source", "statement": "科学 Transformer 模型（如 AlphaFold2）的训练受‘AI 内存墙’制约，现有通用内存优化技术未充分应用于科学架构，需要针对 Evoformer、SE(3)-Transformer 等定制优化，如 AlphaFold2 中的序列维动态并行。", "supportingPaperIds": [ "arxiv:2501.11847" ] }, { "contradictingPaperIds": [], "id": "C11", "maturity": "single_source", "statement": "在社会科学中，将 AI 作为工具在速度、成本、可重复性和可扩展性方面通常优于传统方法，但在效度、数值分析和机制解释方面较弱；LLM 能否替代人类受试者仍是开放问题，目前不能取代社会科学家。", "supportingPaperIds": [ "arxiv:2401.11839" ] }, { "contradictingPaperIds": [], "id": "C12", "maturity": "single_source", "statement": "演化智能视角认为，累积科学发现的关键不仅在于候选解优化，还在于保留搜索历史和失败候选；失败实验和谱系记录可以作为科学证据，用于提炼设计规则和可测试假设。", "supportingPaperIds": [ "arxiv:2607.09025" ] } ], "paperIds": [ "openalex:W4401726605", "openalex:W4412642369", "openalex:W4406828533", "arxiv:2604.27297", "arxiv:2508.14111", "arxiv:2509.02661", "openalex:W3205208140", "openalex:W4323697696", "arxiv:2607.09025", "arxiv:2310.18852", "arxiv:2406.10557", "arxiv:2401.11839", "arxiv:2501.11847", "arxiv:2608.02775", "openalex:W4402901051", "arxiv:2110.01831", "arxiv:2505.03977", "arxiv:2603.28361", "arxiv:2104.08043" ] }, "institutionLandscape": { "coverage": { "availablePaperCount": 5, "status": "available", "totalPaperCount": 19 }, "institutions": [ { "countryCode": "CN", "id": "I3019120426", "name": "China Iron and Steel Research Institute Group", "paperIds": [ "openalex:W4402901051" ], "prominenceScore": 3, "type": "facility" }, { "countryCode": "CN", "id": "I183067930", "name": "Shanghai Jiao Tong University", "paperIds": [ "openalex:W4402901051" ], "prominenceScore": 3, "type": "education" }, { "countryCode": "CN", "id": "I113940042", "name": "Shanghai University", "paperIds": [ "openalex:W4402901051" ], "prominenceScore": 3, "type": "education" }, { "countryCode": "CN", "id": "I4210125098", "name": "Hangzhou Vocational and Technical College", "paperIds": [ "openalex:W4406828533" ], "prominenceScore": 2, "type": "education" }, { "countryCode": "US", "id": "I180949307", "name": "Illinois Institute of Technology", "paperIds": [ "openalex:W4412642369" ], "prominenceScore": 2, "type": "education" }, { "countryCode": "NL", "id": "I4210111138", "name": "Oncode Institute", "paperIds": [ "openalex:W3205208140" ], "prominenceScore": 2, "type": "education" }, { "countryCode": "US", "id": "I219193219", "name": "Purdue University West Lafayette", "paperIds": [ "openalex:W4412642369" ], "prominenceScore": 2, "type": "education" }, { "countryCode": "CN", "id": "I157773358", "name": "Sun Yat-sen University", "paperIds": [ "openalex:W4412642369" ], "prominenceScore": 2, "type": "education" }, { "countryCode": "NL", "id": "I2898336195", "name": "The Netherlands Cancer Institute", "paperIds": [ "openalex:W3205208140" ], "prominenceScore": 2, "type": "healthcare" }, { "countryCode": "GB", "id": "I241749", "name": "University of Cambridge", "paperIds": [ "openalex:W4401726605" ], "prominenceScore": 2, "type": "education" }, { "countryCode": "CN", "id": "I4210105136", "name": "Yangtze River Delta Physics Research Center (China)", "paperIds": [ "openalex:W4406828533" ], "prominenceScore": 2, "type": "company" }, { "countryCode": "CN", "id": "I4210123185", "name": "Zhejiang Lab", "paperIds": [ "openalex:W4406828533" ], "prominenceScore": 2, "type": "facility" }, { "countryCode": "CN", "id": "I76130692", "name": "Zhejiang University", "paperIds": [ "openalex:W4406828533" ], "prominenceScore": 2, "type": "education" } ], "label": "本论文集中的机构显著度", "methodNote": "按入选论文覆盖数和编辑重要性汇总；不代表全球机构排名。" }, "researchCoordinates": [ { "category": "理论与方法", "epistemicStatus": "paper_derived", "evidencePaperIds": [ "openalex:W4412642369", "openalex:W4323697696", "arxiv:2509.02661" ], "name": "物理信息机器学习与偏微分方程求解", "whyItMatters": "用于判断 PINN 等混合建模在科学计算中的潜力和边界，是理解 AI 如何嵌入物理约束的直接入口。" }, { "category": "理论与方法", "epistemicStatus": "paper_derived", "evidencePaperIds": [ "openalex:W4406828533", "openalex:W3205208140" ], "name": "科学序列建模与多模态生物化学语言", "whyItMatters": "AlphaFold 等突破依赖蛋白质/分子序列表示，理解其数据需求和评估缺口需要这一视角。" }, { "category": "数据与评价", "epistemicStatus": "cross_paper_synthesis", "evidencePaperIds": [ "openalex:W4402901051", "openalex:W4401726605", "openalex:W4323697696", "arxiv:2310.18852" ], "name": "AI-ready 数据生态与开放科学机制", "whyItMatters": "数据质量和可获得性被多篇论文识别为采用瓶颈，这一领域知识用于判断某项 AI 声称是否建立在可复用数据之上。" }, { "category": "系统与工程", "epistemicStatus": "cross_paper_synthesis", "evidencePaperIds": [ "arxiv:2508.14111", "arxiv:2604.27297", "arxiv:2608.02775", "arxiv:2607.09025" ], "name": "自主发现系统与闭环实验流程", "whyItMatters": "识别代理科学的存在性证明与可推广能力，区分数字阶段成功和物理实验闭环。" }, { "category": "应用领域", "epistemicStatus": "cross_paper_synthesis", "evidencePaperIds": [ "openalex:W3205208140", "arxiv:2406.10557", "arxiv:2509.02661" ], "name": "可解释性与实验验证标准", "whyItMatters": "帮助判断 AI 输出是否具备科学结论资格，尤其是结构生物学等需要实验仲裁的领域。" }, { "category": "数据与评价", "epistemicStatus": "paper_derived", "evidencePaperIds": [ "arxiv:2104.08043", "arxiv:2505.03977" ], "name": "因果发现与符号回归基准", "whyItMatters": "提供方法选择的反面证据：算法表现高度依赖假设和超参数，不能仅凭单一基准判断优劣。" } ], "schemaVersion": 1, "topicTerms": [ { "category": "主题", "evidencePaperIds": [ "arxiv:2604.27297", "arxiv:2509.02661", "arxiv:2310.18852", "arxiv:2406.10557", "arxiv:2608.02775", "arxiv:2110.01831", "arxiv:2603.28361" ], "term": "cs.AI", "weight": 10 }, { "category": "主题", "evidencePaperIds": [ "openalex:W4401726605", "openalex:W3205208140", "openalex:W4323697696", "openalex:W4402901051" ], "term": "理论与立场", "weight": 10 }, { "category": "主题", "evidencePaperIds": [ "arxiv:2508.14111", "arxiv:2501.11847", "arxiv:2505.03977" ], "term": "cs.LG", "weight": 8 }, { "category": "方法", "evidencePaperIds": [ "openalex:W4406828533" ], "term": "科学大语言模型", "weight": 8 }, { "category": "方法", "evidencePaperIds": [ "openalex:W4412642369" ], "term": "物理信息神经网络", "weight": 7 }, { "category": "主题", "evidencePaperIds": [ "openalex:W4402901051" ], "term": "AI-ready 数据生态", "weight": 6 }, { "category": "主题", "evidencePaperIds": [ "arxiv:2508.14111" ], "term": "代理科学", "weight": 6 }, { "category": "主题", "evidencePaperIds": [ "openalex:W4412642369", "openalex:W4406828533" ], "term": "综述与文献回顾", "weight": 6 }, { "category": "问题", "evidencePaperIds": [ "openalex:W3205208140" ], "term": "实验结构生物学验证", "weight": 5 }, { "category": "方法", "evidencePaperIds": [ "arxiv:2104.08043" ], "term": "时间序列因果发现", "weight": 5 }, { "category": "指标", "evidencePaperIds": [ "arxiv:2505.03977" ], "term": "符号回归基准", "weight": 5 }, { "category": "问题", "evidencePaperIds": [ "arxiv:2501.11847" ], "term": "AI 内存墙", "weight": 4 }, { "category": "主题", "evidencePaperIds": [], "term": "AI for Science", "weight": 3 }, { "category": "主题", "evidencePaperIds": [], "term": "实验验证", "weight": 3 }, { "category": "主题", "evidencePaperIds": [], "term": "数据基础设施", "weight": 3 }, { "category": "主题", "evidencePaperIds": [], "term": "核心共识", "weight": 3 }, { "category": "主题", "evidencePaperIds": [], "term": "自主实验", "weight": 3 }, { "category": "主题", "evidencePaperIds": [], "term": "领域综述", "weight": 3 } ] }, "researchStatus": { "assessment_retry_available": false, "assessment_round": 2, "automatic_supplement_round_limit": 1, "automatic_supplement_rounds_completed": 1, "completion_status": "completed_with_limitations", "continuation_available": true, "evidence_status": "insufficient", "final_high_priority_gap_count": 3, "initial_high_priority_gap_count": 3, "open_gaps": [ { "evidence_paper_ids": [ "openalex:W4412642369", "openalex:W4406828533", "arxiv:2604.27297", "arxiv:2508.14111", "arxiv:2509.02661" ], "id": "GAP-UNIFIED-BENCHMARK", "missing_evidence": "A cross-branch benchmark study that applies multiple competing approaches (PINNs, scientific LLMs, symbolic regression, agents) to the same scientific problems under controlled conditions with common evaluation metrics.", "queries": [ "\"AI for science\" benchmark comparison PINN vs LLM vs symbolic regression", "unified evaluation framework AI for scientific discovery fair comparison" ], "question": "Under a unified benchmark and fair experimental conditions, how do major AI-for-Science branches (e.g., physics-informed networks, scientific LLMs, symbolic regression, autonomous agents) compare on the same scientific tasks?", "rationale": "Current evidence comes from independent reviews and benchmarks within each branch, with no cross-branch comparison using shared datasets and metrics. This prevents identifying which branches are most effective and undermines any reliable conclusion about major research directions and their comparative value.", "severity": "high", "target_evidence": "evaluation" }, { "evidence_paper_ids": [ "openalex:W4401726605", "openalex:W3205208140", "arxiv:2508.14111", "arxiv:2608.02775", "arxiv:2604.27297" ], "id": "GAP-REAL-WORLD-DEPLOYMENT", "missing_evidence": "Large-scale, multi-disciplinary, controlled deployment studies that quantitatively measure discovery speed, cost savings, failure rates, and reproducibility in real laboratory or industrial workflows.", "queries": [ "\"AI for science\" production deployment evaluation research laboratory adoption", "real-world impact of AI on scientific discovery quantitative study" ], "question": "In real scientific production settings (not paper benchmarks), what are the adoption rates, deployment effectiveness, cost-benefit, and failure rates of AI-for-Science systems?", "rationale": "Existing positive evidence is largely based on isolated case studies or small-scale demonstrations, while negative examples (e.g., COVID-19 UK AI underuse) indicate a significant gap. Without systematic deployment data, the claim that AI is transforming science remains speculative and potentially overstated.", "severity": "high", "target_evidence": "scale" }, { "evidence_paper_ids": [ "openalex:W4402901051", "openalex:W4401726605", "arxiv:2509.02661", "arxiv:2310.18852", "arxiv:2608.02775" ], "id": "GAP-CAUSAL-OPEN-DATA", "missing_evidence": "Randomized controlled or quasi-experimental studies that link specific data-sharing policies or AI-ready infrastructure investments to measurable improvements in discovery speed, quality, or reproducibility.", "queries": [ "causal effect of open data on scientific discovery empirical evidence", "AI-ready data infrastructure impact on research productivity study" ], "question": "To what extent do open data sharing, AI-ready data infrastructure, and data governance causally improve scientific discovery output?", "rationale": "Multiple position papers and reviews assert that data and infrastructure are the primary bottlenecks for AI for Science, but this is supported only by normative arguments and illustrative cases. Without causal empirical evidence, this core consensus remains an unvalidated assumption rather than a trustworthy conclusion.", "severity": "high", "target_evidence": "primary" } ], "schema_version": 1, "stop_reason": "automatic_round_limit_reached", "supplementary_completed_count": 3, "supplementary_selected_count": 4 }, "schemaVersion": 3, "scope": { "boundaries": [ "不深入具体算法或优化细节", "不包括 AI for Science 的伦理与政策讨论" ], "missingSources": [ "semantic_scholar", "openalex", "补充检索后证据仍不足：Under a unified benchmark and fair experimental conditions, how do major AI-for-Science branches (e.g., physics-informed networks, scientific LLMs, symbolic regression, autonomous agents) compare on the same scientific tasks?", "补充检索后证据仍不足：In real scientific production settings (not paper benchmarks), what are the adoption rates, deployment effectiveness, cost-benefit, and failure rates of AI-for-Science systems?", "补充检索后证据仍不足：To what extent do open data sharing, AI-ready data infrastructure, and data governance causally improve scientific discovery output?" ], "question": "AI for Science：领域全景与核心共识：关于 AI for Science，现有研究形成了哪些较可信且容易理解的核心结论？主要有哪些研究分支和代表性证据？", "summary": "AI for Science 的可靠共识既包含 AlphaFold 等具体成功，更指向数据就绪、混合建模、可解释性与实验验证共同构成的采用条件，自主发现仍处于存在性证明而非成熟能力。", "topic": "AI for Science：领域全景与核心共识" }, "slug": "research-8715ab7f247f", "subtitle": "不把 AI for Science 写成算法清单，而是沿着数据、模型、实验与制度四条线索，区分当前证据可直接支持的结论与亟待验证的推断。", "supportingEvidence": [ { "abstract": "Artificial intelligence (AI)-driven methods can vastly improve the historically costly drug design process, with various generative models already in widespread use. Generative models for de novo drug design, in particular, focus on the creation of novel biological compounds entirely from scratch, representing a promising future direction. Rapid development in the field, combined with the inherent complexity of the drug design process, creates a difficult landscape for new researchers to enter. In this survey, we organize de novo drug design into two overarching themes: small molecule and protein generation. Within each theme, we identify a variety of subtasks and applications, highlighting important datasets, benchmarks, and model architectures and comparing the performance of top models. We take a broad approach to AI-driven drug design, allowing for both micro-level comparisons of various methods within each subtask and macro-level observations across different fields. We discuss parallel challenges and approaches between the two applications and highlight future directions for AI-driven de novo drug design as a whole. An organized repository of all covered sources is available at https://github.com/gersteinlab/GenAI4Drug.", "evidence_level": "abstract_only", "first_published": "2024-05-23", "limitations": [ "摘要未说明", "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "AI驱动方法可显著改善成本高昂的药物设计过程。", "生成模型专注于从零创建新型生物化合物。", "领域分为小分子和蛋白质生成两大主题，存在平行挑战和方法。" ], "method": "综述性调查，将从头药物设计组织为小分子和蛋白质生成两大主题，识别子任务、数据集、基准和模型架构，并比较性能。", "metrics": { "citation_count": 98, "citation_normalized_percentile": 0.99787111, "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 2, "graph_foundation": 0.0, "graph_frontier": 0.8380359263820629, "graph_reference_out_degree": 1, "graph_structural_role": "frontier", "graph_topology_community": 3, "is_retracted": false, "landscape_relevance_class": "direct", "landscape_relevance_reasons": [ "screening_direct_match" ], "topic_cluster": 2, "topic_cluster_coherence": 0.6257871472780765, "topic_cluster_relation": "direct", "topic_cluster_title": "artificial / intelligence / discovery", "venue": "Briefings in Bioinformatics" }, "paper_id": "openalex:W4400656170", "rank": 21, "relevance": "直接相关，是AI for Science在药物发现领域的代表性综述，提供了领域分支、核心方法和代表性证据。", "screening_reason": "Comprehensive survey of generative AI for drug design, a major AI for Science branch. Directly synthesizes methods, datasets, and benchmarks, providing representative evidence and core findings. High practical relevance for drug discovery.", "source": "openalex", "source_url": "https://doi.org/10.1093/bib/bbae338", "summary_status": "generated", "title": "A survey of generative AI for de novo drug design: new frontiers in molecule and protein generation" }, { "abstract": "In recent years, there has been a dramatic rise in interest in retrosynthesis prediction with artificial intelligence (AI) techniques. Unlike conventional retrosynthesis prediction performed by chemists and by rule-based expert systems, AI-driven retrosynthesis prediction automatically learns chemistry knowledge from off-the-shelf experimental datasets to predict reactions and retrosynthesis routes. This provides an opportunity to address many conventional challenges, including heavy reliance on extensive expertise, the sub-optimality of routes, and prohibitive computational cost. This review describes the current landscape of AI-driven retrosynthesis prediction. We first discuss formal definitions of the retrosynthesis problem and review the outstanding research challenges therein. We then review the related AI techniques and recent progress that enable retrosynthesis prediction. Moreover, we propose a novel landscape that provides a comprehensive categorization of different retrosynthesis prediction components and survey how AI reshapes each component. We conclude by discussing promising areas for future research.", "evidence_level": "abstract_only", "first_published": "2022-08-20", "limitations": [ "摘要未说明", "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "AI驱动的逆合成预测从现成实验数据集中自动学习化学知识。", "该方法解决了依赖广泛专业知识、路线次优和计算成本高的传统挑战。", "提出了一个涵盖逆合成预测组件的综合分类。" ], "method": "综述，形式化定义逆合成预测问题，回顾相关AI技术，提出新的分类框架。", "metrics": { "citation_count": 100, "citation_normalized_percentile": 0.97774475, "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 2, "graph_foundation": 0.0, "graph_frontier": 0.4451672387337635, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "is_retracted": false, "landscape_relevance_class": "direct", "landscape_relevance_reasons": [ "screening_direct_match" ], "topic_cluster": 2, "topic_cluster_coherence": 0.6257871472780765, "topic_cluster_relation": "direct", "topic_cluster_title": "artificial / intelligence / discovery", "venue": "Engineering" }, "paper_id": "openalex:W4292805535", "rank": 22, "relevance": "直接相关，逆合成预测是AI for Chemistry的核心分支，综述提供了方法和进展的可靠综合。", "screening_reason": "Deep review of AI retrosynthesis prediction; representative subfield with credible synthesis of methods and progress.", "source": "openalex", "source_url": "https://doi.org/10.1016/j.eng.2022.04.021", "summary_status": "generated", "title": "Artificial Intelligence for Retrosynthesis Prediction" }, { "abstract": "OBJECTIVES: This study aims to summarize the usage of large language models (LLMs) in the process of creating a scientific review by looking at the methodological papers that describe the use of LLMs in review automation and the review papers that mention they were made with the support of LLMs. MATERIALS AND METHODS: The search was conducted in June 2024 in PubMed, Scopus, Dimensions, and Google Scholar by human reviewers. Screening and extraction process took place in Covidence with the help of LLM add-on based on the OpenAI GPT-4o model. ChatGPT and Scite.ai were used in cleaning the data, generating the code for figures, and drafting the manuscript. RESULTS: Of the 3788 articles retrieved, 172 studies were deemed eligible for the final review. ChatGPT and GPT-based LLM emerged as the most dominant architecture for review automation (n = 126, 73.2%). A significant number of review automation projects were found, but only a limited number of papers (n = 26, 15.1%) were actual reviews that acknowledged LLM usage. Most citations focused on the automation of a particular stage of review, such as Searching for publications (n = 60, 34.9%) and Data extraction (n = 54, 31.4%). When comparing the pooled performance of GPT-based and BERT-based models, the former was better in data extraction with a mean precision of 83.0% (SD = 10.4) and a recall of 86.0% (SD = 9.8). DISCUSSION AND CONCLUSION: Our LLM-assisted systematic review revealed a significant number of research projects related to review automation using LLMs. Despite limitations, such as lower accuracy of extraction for numeric data, we anticipate that LLMs will soon change the way scientific reviews are conducted.", "evidence_level": "abstract_only", "first_published": "2025-05-07", "limitations": [ "数值数据提取的准确性较低", "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "ChatGPT和GPT-based LLM是综述自动化中最主要的架构（占73.2%）。", "大多数项目自动化特定阶段，如搜索（34.9%）和数据提取（31.4%）。", "GPT-based模型在数据提取方面优于BERT-based模型，平均精度83.0%、召回86.0%。" ], "method": "LLM辅助的系统综述，使用GPT-4o辅助筛选和提取，结合四个数据库的文献搜索。", "metrics": { "citation_count": 117, "citation_normalized_percentile": 0.99889989, "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 0, "graph_foundation": 0.0, "graph_frontier": 0.683233370927628, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "is_retracted": false, "landscape_relevance_class": "direct", "landscape_relevance_reasons": [ "screening_direct_match" ], "topic_cluster": 0, "topic_cluster_coherence": 0.5440023663661402, "topic_cluster_relation": "direct", "topic_cluster_title": "science / scientific / intelligence", "venue": "Journal of the American Medical Informatics Association" }, "paper_id": "openalex:W4410157139", "rank": 23, "relevance": "直接相关，展示了LLM作为AI工具在科学文献综述中的应用，是AI for Science的重要分支。", "screening_reason": "LLM-assisted systematic review provides robust evidence on one important branch (AI for literature review), with quantitative performance data; directly contributes to understanding core conclusions in that area.", "source": "openalex", "source_url": "https://doi.org/10.1093/jamia/ocaf063", "summary_status": "generated", "title": "The emergence of large language models as tools in literature reviews: a large language model-assisted systematic review" }, { "abstract": "In recent decades, artificial intelligence has undergone transformative advancements, reshaping diverse sectors such as healthcare, transport, agriculture, energy, and the media. Despite the enthusiasm surrounding AI’s potential, concerns persist about its potential negative impacts, including substantial energy consumption and ethical challenges. This paper critically reviews the evolving landscape of AI sustainability, addressing economic, social, and environmental dimensions. The literature is systematically categorized into “Sustainability of AI” and “AI for Sustainability”, revealing a balanced perspective between the two. The study also identifies a notable trend towards holistic approaches, with a surge in publications and empirical studies since 2019, signaling the field’s maturity. Future research directions emphasize delving into the relatively under-explored economic dimension, aligning with the United Nations’ Sustainable Development Goals (SDGs), and addressing stakeholders’ influence.", "evidence_level": "abstract_only", "first_published": "2024-03-01", "limitations": [ "摘要未说明", "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "文献分为“AI的可持续性”和“AI促进可持续性”两个平衡视角。", "自2019年以来出版物和实证研究激增，表明领域成熟。", "未来研究应关注未被充分探索的经济维度，并与联合国SDGs对接。" ], "method": "批判性综述，系统分类文献为“Sustainability of AI”和“AI for Sustainability”，涵盖经济、社会和环境维度。", "metrics": { "citation_count": 44, "citation_normalized_percentile": 0.99195226, "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 3, "graph_foundation": 0.0, "graph_frontier": 0.5626350620823294, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "is_retracted": false, "landscape_relevance_class": "direct", "landscape_relevance_reasons": [ "screening_direct_match" ], "topic_cluster": 3, "topic_cluster_coherence": 0.5565542221668262, "topic_cluster_relation": "direct", "topic_cluster_title": "artificial / intelligence / review", "venue": "Analytics" }, "paper_id": "openalex:W4392387614", "rank": 24, "relevance": "直接相关，AI for Sustainability是AI for Science的重要分支，提供了系统分类和趋势。", "screening_reason": "Review of AI and sustainability, systematically categorizing literature into 'AI for Sustainability' and 'Sustainability of AI'. Provides clear branches, representative studies, and future directions. Directly relevant to AI for Science and actionable for policy.", "source": "openalex", "source_url": "https://doi.org/10.3390/analytics3010008", "summary_status": "generated", "title": "Artificial Intelligence and Sustainability—A Review" }, { "abstract": "The Vera C. Rubin Observatory's Legacy Survey of Space and Time (LSST) will produce unprecedented volumes of heterogeneous astronomical data (images, catalogs, and alerts) that challenge traditional analysis pipelines. The LSST Dark Energy Science Collaboration (DESC) aims to derive robust constraints on dark energy and dark matter from these data, requiring methods that are statistically powerful, scalable, and operationally reliable. Artificial intelligence and machine learning (AI/ML) are already embedded across DESC science workflows, from photometric redshifts and transient classification to weak lensing inference and cosmological simulations. Yet their utility for precision cosmology hinges on trustworthy uncertainty quantification, robustness to covariate shift and model misspecification, and reproducible integration within scientific pipelines. This white paper surveys the current landscape of AI/ML across DESC's primary cosmological probes and cross-cutting analyses, revealing that the same core methodologies and fundamental challenges recur across disparate science cases. Since progress on these cross-cutting challenges would benefit multiple probes simultaneously, we identify key methodological research priorities, including Bayesian inference at scale, physics-informed methods, validation frameworks, and active learning for discovery. With an eye on emerging techniques, we also explore the potential of the latest foundation model methodologies and LLM-driven agentic AI systems to reshape DESC workflows, provided their deployment is coupled with rigorous evaluation and governance. Finally, we discuss critical software, computing, data infrastructure, and human capital requirements for the successful deployment of these new methodologies, and consider associated risks and opportunities for broader coordination with external actors.", "evidence_level": "abstract_only", "first_published": "2026-01-20", "limitations": [ "摘要未说明", "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "AI/ML已嵌入DESC科学工作流程，包括光度红移、瞬变分类、弱透镜推断和宇宙学模拟。", "相同的核心方法论和基本挑战在不同科学案例中反复出现。", "确定了关键方法论研究优先事项，包括大规模贝叶斯推断、物理信息方法、验证框架和主动学习。" ], "method": "白皮书，调查AI/ML在DESC各主要宇宙学探针和跨领域分析中的现状。", "metrics": { "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 0, "graph_foundation": 0.0, "graph_frontier": 0.55, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "landscape_relevance_class": "direct", "landscape_relevance_reasons": [ "screening_direct_match" ], "topic_cluster": 0, "topic_cluster_coherence": 0.5440023663661402, "topic_cluster_relation": "direct", "topic_cluster_title": "science / scientific / intelligence" }, "paper_id": "arxiv:2601.14235", "rank": 25, "relevance": "直接相关，为AI for Science在天体物理和宇宙学中的应用提供了分支和代表性证据。", "screening_reason": "White paper surveying AI/ML across LSST cosmology probes; identifies core cross-cutting methodologies and challenges.", "source": "arxiv", "source_url": "https://arxiv.org/abs/2601.14235v1", "summary_status": "generated", "title": "Opportunities in AI/ML for the Rubin LSST Dark Energy Science Collaboration" }, { "abstract": "There is an urgent need to improve the infrastructure supporting the reuse of scholarly data. A diverse set of stakeholders-representing academia, industry, funding agencies, and scholarly publishers-have come together to design and jointly endorse a concise and measureable set of principles that we refer to as the FAIR Data Principles. The intent is that these may act as a guideline for those wishing to enhance the reusability of their data holdings. Distinct from peer initiatives that focus on the human scholar, the FAIR Principles put specific emphasis on enhancing the ability of machines to automatically find and use the data, in addition to supporting its reuse by individuals. This Comment is the first formal publication of the FAIR Principles, and includes the rationale behind them, and some exemplar implementations in the community.", "evidence_level": "abstract_only", "first_published": "2016-03-15", "limitations": [ "摘要未说明", "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "FAIR原则旨在增强学术数据的可重用性。", "与关注人类学者的倡议不同，FAIR强调机器的自动查找和使用数据能力。", "提供了原则背后的理由和示例实现。" ], "method": "评论文章，提出FAIR数据原则。", "metrics": { "citation_count": 18312, "citation_normalized_percentile": null, "discovered_via": "reference", "discovery_seed": "openalex:W4412875452", "graph_bridge": 1.0, "graph_citation_in_degree": 5, "graph_community": 1, "graph_foundation": 1.0, "graph_frontier": 0.16513574366602474, "graph_reference_out_degree": 0, "graph_structural_role": "foundation", "graph_topology_community": 1, "is_retracted": false, "landscape_relevance_class": "contextual", "landscape_relevance_reasons": [ "screening_contextual_value" ], "retrieval_lexical_score": 0.011538461538461539, "retrieval_relevance_class": "contextual", "retrieval_relevance_reasons": [ "moderate_semantic_match", "verified_graph_path" ], "retrieval_semantic_score": 0.37432294061814897, "topic_cluster": 1, "topic_cluster_coherence": 0.5138867223246962, "topic_cluster_relation": "direct", "topic_cluster_title": "data / intelligence / materials", "venue": "Scientific Data" }, "paper_id": "openalex:W2302501749", "rank": 26, "relevance": "间接相关，FAIR原则为AI for Science的数据基础设施和机器可读性提供基础，具有上下文价值。", "screening_reason": "FAIR principles are foundational for data sharing but not directly about AI for Science conclusions.", "source": "openalex", "source_url": "https://doi.org/10.1038/sdata.2016.18", "summary_status": "generated", "title": "The FAIR Guiding Principles for scientific data management and stewardship" }, { "abstract": "High throughput experimentation tools, machine learning (ML) methods, and open material databases are radically changing the way new materials are discovered. From the experimentally driven approach in the past, we are moving quickly towards the artificial intelligence (AI) driven approach, realizing the 'inverse design' capabilities that allow the discovery of new materials given the desired properties. This review aims to discuss different principles of AI-driven generative models that are applicable for materials discovery, including different materials representations available for this purpose. We will also highlight specific applications of generative models in designing new catalysts, semiconductors, polymers, or crystals while addressing challenges such as data scarcity, computational cost, interpretability, synthesizability, and dataset biases. Emerging approaches to overcome limitations and integrate AI with experimental workflows will be discussed, including multimodal models, physics informed architectures, and closed-loop discovery systems. This review aims to provide insights for researchers aiming to harness AI's transformative potential in accelerating materials discovery for sustainability, healthcare, and energy innovation.", "evidence_level": "abstract_only", "first_published": "2025-08-05", "limitations": [ "摘要未说明", "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "材料发现正从实验驱动转向AI驱动，实现逆设计。", "生成模型可应用于设计新催化剂、半导体、聚合物和晶体。", "存在数据稀缺、计算成本、可解释性、可合成性和数据集偏差等挑战，新兴方法包括多模态模型、物理信息架构和闭环发现系统。" ], "method": "综述，讨论AI驱动的生成模型在材料发现中的原理、材料表示和应用，并讨论挑战和新兴方法。", "metrics": { "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 2, "graph_foundation": 0.0, "graph_frontier": 0.49500000000000005, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "landscape_relevance_class": "direct", "landscape_relevance_reasons": [ "screening_direct_match" ], "topic_cluster": 2, "topic_cluster_coherence": 0.6257871472780765, "topic_cluster_relation": "direct", "topic_cluster_title": "artificial / intelligence / discovery" }, "paper_id": "arxiv:2508.03278", "rank": 27, "relevance": "直接相关，AI for Materials Discovery是AI for Science的重要分支，提供了核心方法和挑战。", "screening_reason": "Review of generative models in materials discovery; synthesizes core AI approaches, applications, and challenges.", "source": "arxiv", "source_url": "https://arxiv.org/abs/2508.03278v1", "summary_status": "generated", "title": "Artificial Intelligence and Generative Models for Materials Discovery -- A Review" }, { "abstract": "Abstract The United Nations’ 17 Sustainable Development Goals stress the importance of global and local efforts to address inequalities and implement sustainability. Addressing complex, interconnected sustainability challenges requires a systematic, interdisciplinary approach, where technology, AI, and data-driven methods offer potential solutions for optimizing resources, integrating different aspects of sustainability, and informed decision-making. Sustainability research surrounds various local, regional, and global challenges, emphasizing the need to identify emerging areas and gaps where AI and data-driven models play a crucial role. The study performs a comprehensive literature survey and scientometric and semantic analyses, categorizes data-driven methods for sustainability problems, and discusses the sustainable use of AI and big data. The outcomes of the analyses highlight the importance of collaborative and inclusive research that bridges regional differences, the interconnection of AI, technology, and sustainability topics, and the major research themes related to sustainability. It further emphasizes the significance of developing hybrid approaches combining AI, data-driven techniques, and expert knowledge for multi-level, multi-dimensional decision-making. Furthermore, the study recognizes the necessity of addressing ethical concerns and ensuring the sustainable use of AI and big data in sustainability research.", "evidence_level": "abstract_only", "first_published": "2024-05-06", "limitations": [ "摘要未说明", "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "AI、技术和可持续性主题相互联系。", "需要协作和包容性研究来弥合区域差异。", "强调开发结合AI、数据驱动技术和专家知识的混合方法，并解决伦理问题。" ], "method": "综合文献调查和科学计量与语义分析，分类数据驱动方法。", "metrics": { "citation_count": 61, "citation_normalized_percentile": 0.99236043, "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 3, "graph_foundation": 0.0, "graph_frontier": 0.5729593529482199, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "is_retracted": false, "landscape_relevance_class": "direct", "landscape_relevance_reasons": [ "screening_direct_match" ], "topic_cluster": 3, "topic_cluster_coherence": 0.5565542221668262, "topic_cluster_relation": "direct", "topic_cluster_title": "artificial / intelligence / review", "venue": "Journal Of Big Data" }, "paper_id": "openalex:W4396671241", "rank": 28, "relevance": "直接相关，AI for Sustainability是AI for Science的分支，提供了趋势和差距分析。", "screening_reason": "Comprehensive scientometric survey of AI-and-sustainability literature; good branch-level evidence and gap analysis.", "source": "openalex", "source_url": "https://doi.org/10.1186/s40537-024-00912-x", "summary_status": "generated", "title": "Assessing the current landscape of AI and sustainability literature: identifying key trends, addressing gaps and challenges" }, { "abstract": "Drug discovery is adapting to novel technologies such as data science, informatics, and artificial intelligence (AI) to accelerate effective treatment development while reducing costs and animal experiments. AI is transforming drug discovery, as indicated by increasing interest from investors, industrial and academic scientists, and legislators. Successful drug discovery requires optimizing properties related to pharmacodynamics, pharmacokinetics, and clinical outcomes. This review discusses the use of AI in the three pillars of drug discovery: diseases, targets, and therapeutic modalities, with a focus on small molecule drugs. AI technologies, such as generative chemistry, machine learning, and multi-property optimization, have enabled several compounds to enter clinical trials. The scientific community must carefully vet known information to address the reproducibility crisis. The full potential of AI in drug discovery can only be realized with sufficient ground truth and appropriate human intervention at later pipeline stages.", "evidence_level": "abstract_only", "first_published": "2023-07-13", "limitations": [ "摘要未明确说明研究局限，但指出科学界必须仔细审查已知信息以解决可重复性危机，并强调需要足够的基本事实和人工干预，暗示了当前AI方法在数据质量和后期验证方面的限制。", "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "AI正在变革药物发现，投资者、工业与学术科学家及立法者的兴趣日益增加。", "生成化学、机器学习、多属性优化等AI技术已使多种化合物进入临床试验。", "只有具备足够的基本事实并在后期阶段进行适当的人工干预，AI在药物发现中的全部潜力才能实现。" ], "method": "综述性论文，讨论AI在药物发现中的三大支柱（疾病、靶点、治疗方式），重点关注小分子药物，并分析AI技术如生成化学、机器学习、多属性优化。", "metrics": { "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 2, "graph_foundation": 0.0, "graph_frontier": 0.385, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "landscape_relevance_class": "direct", "landscape_relevance_reasons": [ "screening_direct_match" ], "topic_cluster": 2, "topic_cluster_coherence": 0.6257871472780765, "topic_cluster_relation": "direct", "topic_cluster_title": "artificial / intelligence / discovery" }, "paper_id": "arxiv:2307.06521", "rank": 29, "relevance": "直接相关，该综述提供了AI在药物发现这一AI for Science主要分支中的核心结论和代表性临床证据，有助于理解该领域的可信结论。", "screening_reason": "Targeted review assessing AI in drug discovery, offering transparent core conclusions and representative evidence from clinical trials; exemplary survey.", "source": "arxiv", "source_url": "https://arxiv.org/abs/2307.06521v1", "summary_status": "generated", "title": "Artificial Intelligence for Drug Discovery: Are We There Yet?" }, { "abstract": "This paper systematically reviews recent advances in artificial intelligence (AI), with a particular focus on machine learning (ML), across the entire drug discovery pipeline. Due to the inherent complexity, escalating costs, prolonged timelines, and high failure rates of traditional drug discovery methods, there is a critical need to comprehensively understand how AI/ML can be effectively integrated throughout the full process. Currently available literature reviews often narrowly focus on specific phases or methodologies, neglecting the dependence between key stages such as target identification, hit screening, and lead optimization. To bridge this gap, our review provides a detailed and holistic analysis of AI/ML applications across these core phases, highlighting significant methodological advances and their impacts at each stage. We further illustrate the practical impact of these techniques through an in-depth case study focused on hyperuricemia, gout arthritis, and hyperuricemic nephropathy, highlighting real-world successes in molecular target identification and therapeutic candidate discovery. Additionally, we discuss significant challenges facing AI/ML in drug discovery and outline promising future research directions. Ultimately, this review serves as an essential orientation for researchers aiming to leverage AI/ML to overcome existing bottlenecks and accelerate drug discovery.", "evidence_level": "abstract_only", "first_published": "2025-07-04", "limitations": [ "摘要未明确说明具体局限，但指出AI/ML在药物发现中存在重大挑战，意味着现有方法仍有瓶颈。", "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "AI/ML可系统地整合到药物发现整个流程中，弥补了现有综述常忽略关键阶段之间依赖关系的不足。", "基于该案例研究，AI/ML在分子靶点识别和治疗候选发现方面取得了实际成功。", "讨论并指出了AI/ML在药物发现中面临的重大挑战和未来研究方向。" ], "method": "系统综述，全面分析AI/ML在药物发现全流程（靶点识别、命中筛选、先导优化等）中的应用，并通过一个针对高尿酸血症、痛风性关节炎和高尿酸血症肾病的案例研究展示实际影响。", "metrics": { "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 2, "graph_foundation": 0.0, "graph_frontier": 0.49500000000000005, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "landscape_relevance_class": "direct", "landscape_relevance_reasons": [ "screening_direct_match" ], "topic_cluster": 2, "topic_cluster_coherence": 0.6257871472780765, "topic_cluster_relation": "direct", "topic_cluster_title": "artificial / intelligence / discovery" }, "paper_id": "arxiv:2507.03407", "rank": 30, "relevance": "直接相关，该综述提供了药物发现这一核心分支在各个阶段的详细证据和案例，支持AI for Science领域全景中的代表性证据。", "screening_reason": "Comprehensive review of AI in drug discovery, a major AI for Science branch, with detailed evidence and case study.", "source": "arxiv", "source_url": "https://arxiv.org/abs/2507.03407v1", "summary_status": "generated", "title": "Artificial intelligence in drug discovery: A comprehensive review with a case study on hyperuricemia, gout arthritis, and hyperuricemic nephropathy" }, { "abstract": "S ince its introduction, the Checklist for Artificial Intelli- gence in Medical Imaging (CLAIM) has sought to promote complete and consistent reporting of artificial intelligence (AI) science in medical imaging (1).CLAIM has been adopted widely in several medical specialties that involve imaging and AI; as of February 2024, PubMed identified 275 articles that cite the original guideline, and Google Scholar identified 608 citations.CLAIM is one of several reporting guidelines developed to address AI and medical imaging (2).Although not designed as a scoring system, some authors have applied it as such and have found variable adherence among published articles (3-6).Some authors have identified opportunities to improve the guideline, such as separating complex items in the original guideline and accommodating rapidly evolving techniques (7).The CLAIM Steering Committee sought to revise, improve, and formalize the guideline (8).The authors renewed CLAIM's registration with the Enhancing the Quality and Transparency of Health Research (EQUA-TOR) Network, an organization that promotes the use of reporting guidelines to improve health research (https://www.equator-network.org)(9,10).The CLAIM Steering Committee developed and conducted a formal Delphi consensus survey process to review the appropriateness and importance of existing checklist items and to identify new content to reflect current science in AI.The authors recruited 73 volunteers, including physicians from a variety of medical imaging-related specialties, AI scientists, journal editors, and statisticians to form the CLAIM 2024 Update Panel; 72 members completed the two survey rounds and are listed as contributors to this work.To address AI's rapid scientific evolution, this article presents the CLAIM 2024 Update (see Table and see Appendix for downloadable Word document).Based on an expert-panel Delphi process, the guideline's recommendations promote consistent reporting of scientific advances of AI in medical imaging to build trust in published results and enable clinical translation.This guideline serves as an educational tool for both authors and reviewers; it offers a best practice checklist to promote transparency and reproducibility of medical imaging AI research.", "evidence_level": "abstract_only", "first_published": "2024-05-29", "limitations": [ "摘要未明确说明局限，但提到一些作者将其作为评分系统使用时发现已发表文章依从性不一，且原指南的项目被建议进行拆分以应对快速演进的技术。", "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "CLAIM已被广泛采用，截至2024年2月，PubMed有275篇引用，Google Scholar有608篇引用。", "更新版指南基于专家共识，纳入反映当前AI科学的新内容，旨在促进医学影像AI研究的透明度和可重复性，支持临床转化。", "该指南可作为作者和审稿人的教育工具，提供最佳实践清单。" ], "method": "基于德尔菲共识调查过程，由73名志愿者（包括医生、AI科学家、期刊编辑和统计学家）组成专家小组，经过两轮调查更新CLAIM指南。", "metrics": { "citation_count": 405, "citation_normalized_percentile": 0.99990289, "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 4, "graph_foundation": 0.0, "graph_frontier": 0.6335000790313293, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "is_retracted": false, "landscape_relevance_class": "contextual", "landscape_relevance_reasons": [ "screening_contextual_value" ], "topic_cluster": 4, "topic_cluster_coherence": 0.6362756898177276, "topic_cluster_relation": "contextual", "topic_cluster_title": "artificial / intelligence / 2024", "venue": "Radiology Artificial Intelligence" }, "paper_id": "openalex:W4399125735", "rank": 31, "relevance": "属于AI for Science在医学影像这一具体应用分支的标准化工作，提供了该领域报告规范的代表性证据，但对领域全景的核心共识贡献有限。", "screening_reason": "Reporting guideline for AI in medical imaging, a specific application area, but not field-wide conclusions.", "source": "openalex", "source_url": "https://doi.org/10.1148/ryai.240300", "summary_status": "generated", "title": "Checklist for Artificial Intelligence in Medical Imaging (CLAIM): 2024 Update" }, { "abstract": "It is increasingly recognised that advances in artificial intelligence could have large and long-lasting impacts on society. However, what form those impacts will take, just how large and long-lasting they will be, and whether they will ultimately be positive or negative for humanity, is far from clear. Based on surveying literature on the societal impacts of AI, we identify and discuss five potential long-term impacts of AI: how AI could lead to long-term changes in science, cooperation, power, epistemics, and values. We review the state of existing research in each of these areas and highlight priority questions for future research.", "evidence_level": "abstract_only", "first_published": "2022-06-22", "limitations": [ "摘要未明确说明局限，但指出AI长期影响的形态、大小和方向尚不清晰，这本身构成研究不确定性的限制。", "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "AI可能对科学、合作、权力、认识论和价值观产生长期影响。", "这些影响的具体形式、幅度和最终正负尚不明确。", "综述了各领域现有研究，并提出了未来研究的优先问题。" ], "method": "基于对AI社会影响文献的广泛调查，识别并讨论了AI的五种潜在长期影响：科学、合作、权力、认识论和价值观。", "metrics": { "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 1, "graph_foundation": 0.0, "graph_frontier": 0.33, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "landscape_relevance_class": "direct", "landscape_relevance_reasons": [ "screening_direct_match" ], "topic_cluster": 1, "topic_cluster_coherence": 0.5138867223246962, "topic_cluster_relation": "direct", "topic_cluster_title": "data / intelligence / materials" }, "paper_id": "arxiv:2206.11076", "rank": 32, "relevance": "直接相关，该综述专门论述了AI对科学的长期影响，是AI for Science领域全景中关于核心共识和未来研究重点的重要文献。", "screening_reason": "Comprehensive survey of AI's long-term impacts, with a dedicated section on science; synthesizes key findings and research priorities.", "source": "arxiv", "source_url": "https://arxiv.org/abs/2206.11076v1", "summary_status": "generated", "title": "A Survey of the Potential Long-term Impacts of AI" }, { "abstract": "Brain-inspired computing (BIC) is an emerging research field that aims to build fundamental theories, models, hardware architectures, and application systems toward more general artificial intelligence (AI) by learning from the information processing mechanisms or structures/functions of biological nervous systems. It is regarded as one of the most promising research directions for future intelligent computing in the post-Moore era. In the past few years, various new schemes in this field have sprung up to explore more general AI. These works are quite divergent in the aspects of modeling/algorithm, software tool, hardware platform, and benchmark data since BIC is an interdisciplinary field that consists of many different domains, including computational neuroscience, AI, computer science, statistical physics, material science, and microelectronics. This situation greatly impedes researchers from obtaining a clear picture and getting started in the right way. Hence, there is an urgent requirement to do a comprehensive survey in this field to help correctly recognize and analyze such bewildering methodologies. What are the key issues to enhance the development of BIC? What roles do the current mainstream technologies play in the general framework of BIC? Which techniques are truly useful in real-world applications? These questions largely remain open. To address the above issues, in this survey, we first clarify the biggest challenge of BIC: how can AI models benefit from the recent advancements in computational neuroscience? With this challenge in mind, we will focus on discussing the concept of BIC and summarize four components of BIC infrastructure development: 1) modeling/algorithm; 2) hardware platform; 3) software tool; and 4) benchmark data. For each component, we will summarize its recent progress, main challenges to resolve, and future trends. Based on these studies, we present a general framework for the real-world applications of BIC systems, which is promising to benefit both AI and brain science. Finally, we claim that it is extremely important to build a research ecology to promote prosperity continuously in this field.", "evidence_level": "abstract_only", "first_published": "2024-06-01", "limitations": [ "摘要未明确说明研究局限，但指出关键问题（如AI模型如何受益于计算神经科学进展）在很大程度上仍未解决，表明该领域成熟度不足。", "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "脑启发计算被认为是后摩尔时代未来智能计算最有前途的研究方向之一。", "该领域方法多样，跨多个学科，亟需全面调查来帮助研究者正确认识和分析方法。", "构建研究生态对持续促进该领域繁荣至关重要。" ], "method": "系统综述，从建模/算法、硬件平台、软件工具和基准数据四个组成部分总结脑启发计算（BIC）的进展、挑战和未来趋势，并提出通用应用框架。", "metrics": { "citation_count": 75, "citation_normalized_percentile": 0.9969346, "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 1, "graph_foundation": 0.0, "graph_frontier": 0.5795184767431464, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "is_retracted": false, "landscape_relevance_class": "direct", "landscape_relevance_reasons": [ "screening_direct_match" ], "topic_cluster": 1, "topic_cluster_coherence": 0.5138867223246962, "topic_cluster_relation": "direct", "topic_cluster_title": "data / intelligence / materials", "venue": "Proceedings of the IEEE" }, "paper_id": "openalex:W4401596736", "rank": 33, "relevance": "属于AI for Science中AI技术演进的相关分支，其目标是实现更通用的AI，对科学研究有间接但重要的支撑作用，但缺乏直接科学应用结论。", "screening_reason": "Systematic survey of brain-inspired computing, an AI research branch with potential for scientific discovery. Covers modeling, hardware, software, and benchmarks, but is more focused on AI techniques than direct science applications. High-quality synthesis.", "source": "openalex", "source_url": "https://doi.org/10.1109/jproc.2024.3429360", "summary_status": "generated", "title": "Brain-Inspired Computing: A Systematic Survey and Future Trends" }, { "abstract": "Abstract The use of Artificial Intelligence (AI) in education is transforming various dimensions of the education system, such as instructional practices, assessment strategies, and administrative processes. It also plays an active role in the progression of science education. This systematic review attempts to render an inherent understanding of the evidence-based interaction between AI and science education. Specifically, this study offers a consolidated analysis of AI’s impact on students’ learning outcomes, contexts of its adoption, students’ and teachers’ perceptions about its use, and the challenges of its use within science education. The present study followed the PRISMA guidelines to review empirical papers published from 2014 to 2023. In total, 74 records met the eligibility for this systematic study. Previous research provides evidence of AI integration into a variety of fields in physical and natural sciences in many countries across the globe. The results revealed that AI-powered tools are integrated into science education to achieve various pedagogical benefits, including enhancing the learning environment, creating quizzes, assessing students’ work, and predicting their academic performance. The findings from this paper have implications for teachers, educational administrators, and policymakers.", "evidence_level": "abstract_only", "first_published": "2024-06-27", "limitations": [ "摘要未明确说明研究局限，但系统综述可能受限于纳入研究的质量和范围，具体限制未提及。", "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "AI工具已被整合到多个国家和物理/自然科学的多个领域中，用于科学教育。", "AI在科学教育中实现了多种教学益处，包括增强学习环境、创建测验、评估学生作业和预测学业表现。", "研究为教师、教育管理者和政策制定者提供了启示。" ], "method": "遵循PRISMA指南的系统综述，筛选2014至2023年间发表的实证研究，最终纳入74篇，分析AI在科学教育中的影响、应用背景、师生认知及挑战。", "metrics": { "citation_count": 287, "citation_normalized_percentile": 0.98888889, "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 3, "graph_foundation": 0.0, "graph_frontier": 0.6224373744500663, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "is_retracted": false, "landscape_relevance_class": "contextual", "landscape_relevance_reasons": [ "screening_contextual_value" ], "topic_cluster": 3, "topic_cluster_coherence": 0.5565542221668262, "topic_cluster_relation": "direct", "topic_cluster_title": "artificial / intelligence / review", "venue": "Research in Science Education" }, "paper_id": "openalex:W4400098474", "rank": 35, "relevance": "该综述涉及AI在科学教育中的应用，属于AI for Science的外围分支，与领域研究核心结论关系较弱，但提供了教育应用这一维度的证据。", "screening_reason": "Systematic review on AI in science education; well-evidenced but addresses a peripheral branch relative to the AI for Science research focus.", "source": "openalex", "source_url": "https://doi.org/10.1007/s11165-024-10176-3", "summary_status": "generated", "title": "Exploring the Impact of Artificial Intelligence in Teaching and Learning of Science: A Systematic Review of Empirical Research" }, { "abstract": "We present a survey of ways in which existing scientific knowledge are included when constructing models with neural networks. The inclusion of domain-knowledge is of special interest not just to constructing scientific assistants, but also, many other areas that involve understanding data using human-machine collaboration. In many such instances, machine-based model construction may benefit significantly from being provided with human-knowledge of the domain encoded in a sufficiently precise form. This paper examines the inclusion of domain-knowledge by means of changes to: the input, the loss-function, and the architecture of deep networks. The categorisation is for ease of exposition: in practice we expect a combination of such changes will be employed. In each category, we describe techniques that have been shown to yield significant changes in the performance of deep neural networks.", "evidence_level": "abstract_only", "first_published": "2022-01-20", "limitations": [ "摘要未明确说明研究局限，但提到分类是为了便于说明，实际中可能组合使用，没有讨论各种技术的相对优劣或失败案例。", "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "领域知识可以通过修改输入、损失函数和架构三种途径纳入深度神经网络。", "这些技术已被证明能显著改变深度神经网络的性能。", "实际应用中往往混合使用多种技术，而非单一方式。" ], "method": "综述性研究，通过输入、损失函数和网络架构三个类别，系统总结了将科学领域知识纳入深度神经网络的多种技术，并预期实践中会组合使用。", "metrics": { "citation_count": 181, "citation_normalized_percentile": 0.99475071, "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 1, "graph_foundation": 0.0, "graph_frontier": 0.45986249834985393, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "is_retracted": false, "landscape_relevance_class": "direct", "landscape_relevance_reasons": [ "screening_direct_match" ], "topic_cluster": 1, "topic_cluster_coherence": 0.5138867223246962, "topic_cluster_relation": "direct", "topic_cluster_title": "data / intelligence / materials", "venue": "Scientific Reports" }, "paper_id": "openalex:W4205511145", "rank": 36, "relevance": "直接相关，该综述聚焦AI for Science中知识融合这一核心主题，提供了将先验科学知识与深度学习结合的技术路径和代表性证据。", "screening_reason": "Review of domain-knowledge integration into deep networks, a key AI4Science branch; credible but not a field-wide synthesis.", "source": "openalex", "source_url": "https://doi.org/10.1038/s41598-021-04590-0", "summary_status": "generated", "title": "A review of some techniques for inclusion of domain-knowledge into deep neural networks" }, { "abstract": "This literature research had two main objectives. The first objective was to quantify how frequently artificial intelligence (AI) was utilized in dental literature from 2011 until 2021. The second objective was to distinguish the focus of such publications; in particular, dental field and topic. The main inclusion criterium was an original article or review in English focused on dental utilization of AI. All other types of publications or non-dental or non-AI-focused were excluded. The information sources were Web of Science, PubMed, Scopus, and Google Scholar, queried on 19 April 2022. The search string was \"artificial intelligence\" AND (dental OR dentistry OR tooth OR teeth OR dentofacial OR maxillofacial OR orofacial OR orthodontics OR endodontics OR periodontics OR prosthodontics). Following the removal of duplicates, all remaining publications were returned by searches and were screened by three independent operators to minimize the risk of bias. The analysis of 2011-2021 publications identified 4413 records, from which 1497 were finally selected and calculated according to the year of publication. The results confirmed a historically unprecedented boom in AI dental publications, with an average increase of 21.6% per year over the last decade and a 34.9% increase per year over the last 5 years. In the achievement of the second objective, qualitative assessment of dental AI publications since 2021 identified 1717 records, with 497 papers finally selected. The results of this assessment indicated the relative proportions of focal topics, as follows: radiology 26.36%, orthodontics 18.31%, general scope 17.10%, restorative 12.09%, surgery 11.87% and education 5.63%. The review confirms that the current use of artificial intelligence in dentistry is concentrated mainly around the evaluation of digital diagnostic methods, especially radiology; however, its implementation is expected to gradually penetrate all parts of the profession.", "evidence_level": "abstract_only", "first_published": "2022-07-08", "limitations": [ "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "摘要原文（未完成模型归纳）：This literature research had two main objectives. The first objective was to quantify how frequently artificial intelligence (AI) was utilized in dental literature from 2011 until 2021. The second objective was to distinguish the focus of such publications; in particular, dental field and topic. The main inclusion criterium was an original article or review in English focused on dental utilization of AI. All other types of publications or non-dental or non-AI-focused were excluded. The information sources were Web of Science, PubMed, Scopus, and Google Scholar, queried on 19 April 2022. The search string was \"artificial intelligence\" AND (dental OR dentistry OR tooth OR teeth OR dentofacial OR maxillofacial OR orofacial OR orthodontics OR endodontics OR periodontics OR prosthodontics). Following the removal of duplicates, all remaining publications were returned by searches and were screened by three independent operators to minimize the risk of bias. The analysis of 2011" ], "method": "摘要未明确说明研究方法。", "metrics": { "citation_count": 189, "citation_normalized_percentile": 0.99844237, "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 5, "graph_foundation": 0.0, "graph_frontier": 0.46093596835774586, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "is_retracted": false, "landscape_relevance_class": "direct", "landscape_relevance_reasons": [ "screening_direct_match" ], "topic_cluster": 5, "topic_cluster_coherence": 0.6593335309716388, "topic_cluster_relation": "direct", "topic_cluster_title": "artificial / intelligence / covid-19", "venue": "Healthcare" }, "paper_id": "openalex:W4284887878", "rank": 37, "relevance": "Systematic review of AI in dentistry, quantifying publication trends and mapping subfields; robust evidence but domain-specific, moderately representative of AI for science.", "screening_reason": "Systematic review of AI in dentistry, quantifying publication trends and mapping subfields; robust evidence but domain-specific, moderately representative of AI for science.", "source": "openalex", "source_url": "https://doi.org/10.3390/healthcare10071269", "summary_status": "fallback", "title": "Where Is the Artificial Intelligence Applied in Dentistry? Systematic Review and Literature Analysis" }, { "abstract": "Artificial intelligence (AI) has emerged as a transformative and versatile tool, breaking new frontiers across scientific domains. Among its most promising applications, AI research is blossoming in concrete science and engineering, where it has offered new insights towards mixture design optimization and service life prediction of cementitious systems. This chapter aims to uncover the main research interests and knowledge structure of the existing literature on AI for concrete materials. To begin with, a total of 389 journal articles published from 1990 to 2020 were retrieved from the Web of Science. Scientometric tools such as keyword co-occurrence analysis and documentation co-citation analysis were adopted to quantify features and characteristics of the research field. The findings bring to light pressing questions in data-driven concrete research and suggest future opportunities for the concrete community to fully utilize the capabilities of AI techniques.", "evidence_level": "abstract_only", "first_published": "2022-09-17", "limitations": [ "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "摘要原文（未完成模型归纳）：Artificial intelligence (AI) has emerged as a transformative and versatile tool, breaking new frontiers across scientific domains. Among its most promising applications, AI research is blossoming in concrete science and engineering, where it has offered new insights towards mixture design optimization and service life prediction of cementitious systems. This chapter aims to uncover the main research interests and knowledge structure of the existing literature on AI for concrete materials. To begin with, a total of 389 journal articles published from 1990 to 2020 were retrieved from the Web of Science. Scientometric tools such as keyword co-occurrence analysis and documentation co-citation analysis were adopted to quantify features and characteristics of the research field. The findings bring to light pressing questions in data-driven concrete research and suggest future opportunities for the concrete community to fully utilize the capabilities of AI techniques." ], "method": "摘要未明确说明研究方法。", "metrics": { "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 3, "graph_foundation": 0.0, "graph_frontier": 0.33, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "landscape_relevance_class": "direct", "landscape_relevance_reasons": [ "screening_direct_match" ], "topic_cluster": 3, "topic_cluster_coherence": 0.5565542221668262, "topic_cluster_relation": "direct", "topic_cluster_title": "artificial / intelligence / review" }, "paper_id": "arxiv:2209.09636", "rank": 38, "relevance": "Scientometric review of AI in concrete materials offers a quantified panorama and research trends for a key branch, aligning with the request for domain panoramas.", "screening_reason": "Scientometric review of AI in concrete materials offers a quantified panorama and research trends for a key branch, aligning with the request for domain panoramas.", "source": "arxiv", "source_url": "https://arxiv.org/abs/2209.09636v1", "summary_status": "fallback", "title": "Artificial Intelligence in Concrete Materials: A Scientometric View" }, { "abstract": "BACKGROUND: COVID-19 was first discovered in December 2019 and has since evolved into a pandemic. OBJECTIVE: To address this global health crisis, artificial intelligence (AI) has been deployed at various levels of the health care system. However, AI has both potential benefits and limitations. We therefore conducted a review of AI applications for COVID-19. METHODS: We performed an extensive search of the PubMed and EMBASE databases for COVID-19-related English-language studies published between December 1, 2019, and March 31, 2020. We supplemented the database search with reference list checks. A thematic analysis and narrative review of AI applications for COVID-19 was conducted. RESULTS: In total, 11 papers were included for review. AI was applied to COVID-19 in four areas: diagnosis, public health, clinical decision making, and therapeutics. We identified several limitations including insufficient data, omission of multimodal methods of AI-based assessment, delay in realization of benefits, poor internal/external validation, inability to be used by laypersons, inability to be used in resource-poor settings, presence of ethical pitfalls, and presence of legal barriers. AI could potentially be explored in four other areas: surveillance, combination with big data, operation of other core clinical services, and management of patients with COVID-19. CONCLUSIONS: In view of the continuing increase in the number of cases, and given that multiple waves of infections may occur, there is a need for effective methods to help control the COVID-19 pandemic. Despite its shortcomings, AI holds the potential to greatly augment existing human efforts, which may otherwise be overwhelmed by high patient numbers.", "evidence_level": "abstract_only", "first_published": "2020-09-15", "limitations": [ "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "摘要原文（未完成模型归纳）：BACKGROUND: COVID-19 was first discovered in December 2019 and has since evolved into a pandemic. OBJECTIVE: To address this global health crisis, artificial intelligence (AI) has been deployed at various levels of the health care system. However, AI has both potential benefits and limitations. We therefore conducted a review of AI applications for COVID-19. METHODS: We performed an extensive search of the PubMed and EMBASE databases for COVID-19-related English-language studies published between December 1, 2019, and March 31, 2020. We supplemented the database search with reference list checks. A thematic analysis and narrative review of AI applications for COVID-19 was conducted. RESULTS: In total, 11 papers were included for review. AI was applied to COVID-19 in four areas: diagnosis, public health, clinical decision making, and therapeutics. We identified several limitations including insufficient data, omission of multimodal methods of AI-based assessment, delay in " ], "method": "摘要未明确说明研究方法。", "metrics": { "citation_count": 162, "citation_normalized_percentile": 0.98891375, "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 5, "graph_foundation": 0.0, "graph_frontier": 0.3274285052823043, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "is_retracted": false, "landscape_relevance_class": "direct", "landscape_relevance_reasons": [ "screening_direct_match" ], "topic_cluster": 5, "topic_cluster_coherence": 0.6593335309716388, "topic_cluster_relation": "direct", "topic_cluster_title": "artificial / intelligence / covid-19", "venue": "Journal of Medical Internet Research" }, "paper_id": "openalex:W3087747282", "rank": 41, "relevance": "Rapid review of AI applications for COVID-19. Offers representative evidence in a specific branch (pandemic response) and identifies limitations, but lacks broad synthesis across AI for Science. Useful for crisis-related AI deployment.", "screening_reason": "Rapid review of AI applications for COVID-19. Offers representative evidence in a specific branch (pandemic response) and identifies limitations, but lacks broad synthesis across AI for Science. Useful for crisis-related AI deployment.", "source": "openalex", "source_url": "https://doi.org/10.2196/21476", "summary_status": "fallback", "title": "Artificial Intelligence for COVID-19: Rapid Review" }, { "abstract": "Diffusion models have become a new SOTA generative modeling method in various fields, for which there are multiple survey works that provide an overall survey. With the number of articles on diffusion models increasing exponentially in the past few years, there is an increasing need for surveys of diffusion models on specific fields. In this work, we are committed to conducting a survey on the graph diffusion models. Even though our focus is to cover the progress of diffusion models in graphs, we first briefly summarize how other generative modeling methods are used for graphs. After that, we introduce the mechanism of diffusion models in various forms, which facilitates the discussion on the graph diffusion models. The applications of graph diffusion models mainly fall into the category of AI-generated content (AIGC) in science, for which we mainly focus on how graph diffusion models are utilized for generating molecules and proteins but also cover other cases, including materials design. Moreover, we discuss the issue of evaluating diffusion models in the graph domain and the existing challenges.", "evidence_level": "abstract_only", "first_published": "2023-01-01", "limitations": [ "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "摘要原文（未完成模型归纳）：Diffusion models have become a new SOTA generative modeling method in various fields, for which there are multiple survey works that provide an overall survey. With the number of articles on diffusion models increasing exponentially in the past few years, there is an increasing need for surveys of diffusion models on specific fields. In this work, we are committed to conducting a survey on the graph diffusion models. Even though our focus is to cover the progress of diffusion models in graphs, we first briefly summarize how other generative modeling methods are used for graphs. After that, we introduce the mechanism of diffusion models in various forms, which facilitates the discussion on the graph diffusion models. The applications of graph diffusion models mainly fall into the category of AI-generated content (AIGC) in science, for which we mainly focus on how graph diffusion models are utilized for generating molecules and proteins but also cover other cases, including" ], "method": "摘要未明确说明研究方法。", "metrics": { "citation_count": 28, "citation_normalized_percentile": null, "graph_bridge": 0.0, "graph_citation_in_degree": 1, "graph_community": 2, "graph_foundation": 0.2, "graph_frontier": 0.4789468421834329, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 3, "is_retracted": false, "landscape_relevance_class": "direct", "landscape_relevance_reasons": [ "screening_direct_match" ], "topic_cluster": 2, "topic_cluster_coherence": 0.6257871472780765, "topic_cluster_relation": "direct", "topic_cluster_title": "artificial / intelligence / discovery", "venue": "arXiv (Cornell University)" }, "paper_id": "openalex:W4362655746", "rank": 42, "relevance": "Survey of graph diffusion models for molecule/protein/material generation; directly relevant, synthesizes evidence for a major AI4Science branch.", "screening_reason": "Survey of graph diffusion models for molecule/protein/material generation; directly relevant, synthesizes evidence for a major AI4Science branch.", "source": "openalex", "source_url": "http://arxiv.org/abs/2304.01565", "summary_status": "fallback", "title": "A Survey on Graph Diffusion Models: Generative AI in Science for Molecule, Protein and Material" }, { "abstract": "“To improve is to change; to be perfect is to change often.” ― Winston S. Churchill. In a time and age where technology is evolving at a sometimes incomprehensibly rapid pace, the liver community must grow with its challenges and adjust our practices to transformative influences on our science and practice. The Editorial Board of the Journal of Hepatology has previously responded to novel developments in Artificial Intelligence (AI) by including experts in the field into the editorial board. Publications utilizing AI technology are no longer uncommon in our journal and have become among the most highly debated and possibly practice-changing papers across all disciplines united by our focus on liver disease. As AI is rapidly evolving, this expert paper will focus on large language models and their possible impact on our research practice and clinical outlook, outlining both challenges and opportunities in the field. Generative AI creates new content, such as text, images, or music, based on patterns in the data it has been trained on. If Generative AI utilizes Large Language Models, it involves training on extensive text datasets and employing AI models with a large number of model parameters. ChatGPT (General Pretrained Transformers) is a popular example of a Generative AI that utilizes Large Language Models. It was developed by OpenAI, version 3.5 released in November 2022, then updated to its superior version 4.0 – called GPT-4 - in March 2023 and version 5 expected to be released in the near future. Figure 1 illustrates the terminology within the overarching concept of Artificial Intelligence (AI). The term AI was coined by McCarthy in the 1950s [1]Shannon CE, McCarthy J, Ashby WR. Automata studies. Princeton University Press. 1956;Google Scholar and describes a system that can mimic human behavior. This can be realized via expert-driven rule-based systems or by data-driven training via Machine Learning. The latter is capable of accomplishing classification tasks based on predetermined categories and features, e.g. recognizing patterns in data that have been previously labeled by an expert on a training data set. Deep Learning architectures are a specific subtype of Machine Learning that utilize an Artificial Neural Network design characterized by an input layer for data intake, multiple hidden layers of data analyses, and finally an output layer that characterizes the original data according to either a predetermined classification task or outlines freshly identified patterns in data that have not yet been identified by the human observer. While classical supervised Machine or Deep Learning requires labeled training data and provides a reduced fixed output like a numerical prediction (e.g. “this patient with Hepatocellular carcinoma will benefit from immunotherapy with 90% certainty”), unsupervised Neural Networks do not require labelled data and are able to detect previously unknown patterns in data. An advancement is semi-supervised Generative AI, which is trained on unlabeled data and then fine-tuned for specific supervised tasks. It can create more complex output based on input prompts and – in its most advanced version – may generate entirely new data contexts. For instance, ChatGPT utilizes transformer Neural Networks, which are pre-trained on unlabeled large text corpora to acquire a comprehensive understanding of language patterns. At this stage it is called a Large Langue Model. The model is then further fine-tuned in order to answer user prompts. The principal approach of pre-training and finetuning can be adapted for different tasks beyond text data. For instance, new realistic images can be generated based on user-prompts, as showcased by Dall-E, another creation of OpenAI. [2]Marcus G, Davis E, Aaronson S. A very preliminary analysis of DALL-E 2 [Internet]. arXiv; 2022 [cited 2023 Jun 24]. Available from: http://arxiv.org/abs/2204.13807Google Scholar. Based on the aforementioned recent advancement of Artificial Neural Networks, which are powered by increasing hardware resources and the immense amount of available human-written text that serves as training basis, ChatGPT generates human-like text better than most other Natural Language Processing tools [3]Deng J. Lin Y. The Benefits and Challenges of ChatGPT: An Overview.Frontiers in Computing and Intelligent Systems. 2022; 2: 81-83Crossref Google Scholar. It is highly effective in generating well-formulated text, which will be disruptive in the way we are going to create text, including academic writing in the biomedical domain. Users can type in raw text containing bullet points or subheadings, which can be processed by ChatGPT to coherent text, allowing the user to carefully refine and thereby creating well-written and meaningful text. In particular, this will be helpful for non-native speakers by enabling technology-based democratization of language skills. The advanced processing of language syntax has the ability to improve upon existing language input and semantically add to the initial prompt while generating explanations, references or – in the best case scenario – fact-check the existing reasoning as offered by the primary end user. In addition to the advantages of promising text completion or supporting writing skills, some early applications have been systematically tested for specific tasks. These include supporting computer programmers in advanced tasks such as converting or generating programming source code, proof-reading code and bug-fixing ,[4]Surameery N.M.S. Shakor M.Y. Use Chat GPT to Solve Programming Bugs.International Journal of Information Technology & Computer Engineering (IJITC). 2023 Jan 28; 3 (ISSN : 2455-5290): 17-22Crossref Google Scholar, classifying hate-speeches on twitter [5]Huang F, Kwak H, An J. Is ChatGPT better than Human Annotators? Potential and Limitations of ChatGPT in Explaining Implicit Hate Speech. In: Companion Proceedings of the ACM Web Conference 2023 [Internet]. New York, ", "evidence_level": "abstract_only", "first_published": "2023-08-06", "limitations": [ "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "摘要原文（未完成模型归纳）：“To improve is to change; to be perfect is to change often.” ― Winston S. Churchill. In a time and age where technology is evolving at a sometimes incomprehensibly rapid pace, the liver community must grow with its challenges and adjust our practices to transformative influences on our science and practice. The Editorial Board of the Journal of Hepatology has previously responded to novel developments in Artificial Intelligence (AI) by including experts in the field into the editorial board. Publications utilizing AI technology are no longer uncommon in our journal and have become among the most highly debated and possibly practice-changing papers across all disciplines united by our focus on liver disease. As AI is rapidly evolving, this expert paper will focus on large language models and their possible impact on our research practice and clinical outlook, outlining both challenges and opportunities in the field. Generative AI creates new content, such as text, images, " ], "method": "摘要未明确说明研究方法。", "metrics": { "citation_count": 93, "citation_normalized_percentile": 0.92419038, "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 1, "graph_foundation": 0.0, "graph_frontier": 0.5117569644762838, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "is_retracted": false, "landscape_relevance_class": "contextual", "landscape_relevance_reasons": [ "screening_contextual_value" ], "topic_cluster": 1, "topic_cluster_coherence": 0.5138867223246962, "topic_cluster_relation": "direct", "topic_cluster_title": "data / intelligence / materials", "venue": "Journal of Hepatology" }, "paper_id": "openalex:W4385605341", "rank": 44, "relevance": "Expert perspective on LLMs in science and healthcare; highlights opportunities/risks but is editorial, not synthetic survey.", "screening_reason": "Expert perspective on LLMs in science and healthcare; highlights opportunities/risks but is editorial, not synthetic survey.", "source": "openalex", "source_url": "https://doi.org/10.1016/j.jhep.2023.07.028", "summary_status": "fallback", "title": "ChatGPT: The transformative influence of generative AI on science and healthcare" }, { "abstract": "Large language models (LLMs) leverage chain-of-thought (CoT) techniques to tackle complex problems, representing a transformative breakthrough in artificial intelligence (AI). However, their reasoning capabilities have primarily been demonstrated in solving math and coding problems, leaving their potential for domain-specific applications-such as battery discovery-largely unexplored. Inspired by the idea that reasoning mirrors a form of guided search, we introduce ChatBattery, a novel agentic framework that integrates domain knowledge to steer LLMs toward more effective reasoning in materials design. Using ChatBattery, we successfully identify, synthesize, and characterize three novel lithium-ion battery cathode materials, which achieve practical capacity improvements of 28.8%, 25.2%, and 18.5%, respectively, over the widely used cathode material, LiNi0.8Mn0.1Co0.1O2 (NMC811). Beyond this discovery, ChatBattery paves a new path by showing a successful LLM-driven and reasoning-based platform for battery materials invention. This complete AI-driven cycle-from design to synthesis to characterization-demonstrates the transformative potential of AI-driven reasoning in revolutionizing materials discovery.", "evidence_level": "abstract_only", "first_published": "2025-07-21", "limitations": [ "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "摘要原文（未完成模型归纳）：Large language models (LLMs) leverage chain-of-thought (CoT) techniques to tackle complex problems, representing a transformative breakthrough in artificial intelligence (AI). However, their reasoning capabilities have primarily been demonstrated in solving math and coding problems, leaving their potential for domain-specific applications-such as battery discovery-largely unexplored. Inspired by the idea that reasoning mirrors a form of guided search, we introduce ChatBattery, a novel agentic framework that integrates domain knowledge to steer LLMs toward more effective reasoning in materials design. Using ChatBattery, we successfully identify, synthesize, and characterize three novel lithium-ion battery cathode materials, which achieve practical capacity improvements of 28.8%, 25.2%, and 18.5%, respectively, over the widely used cathode material, LiNi0.8Mn0.1Co0.1O2 (NMC811). Beyond this discovery, ChatBattery paves a new path by showing a successful LLM-driven and reaso" ], "method": "摘要未明确说明研究方法。", "metrics": { "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 0, "graph_foundation": 0.0, "graph_frontier": 0.49500000000000005, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "landscape_relevance_class": "contextual", "landscape_relevance_reasons": [ "screening_contextual_value" ], "topic_cluster": 0, "topic_cluster_coherence": 0.5440023663661402, "topic_cluster_relation": "direct", "topic_cluster_title": "science / scientific / intelligence" }, "paper_id": "arxiv:2507.16110", "rank": 45, "relevance": "Demonstrates a complete AI-driven materials discovery cycle, serving as representative evidence for one branch, but does not synthesize core conclusions across the field.", "screening_reason": "Demonstrates a complete AI-driven materials discovery cycle, serving as representative evidence for one branch, but does not synthesize core conclusions across the field.", "source": "arxiv", "source_url": "https://arxiv.org/abs/2507.16110v1", "summary_status": "fallback", "title": "Expert-Guided LLM Reasoning for Battery Discovery: From AI-Driven Hypothesis to Synthesis and Characterization" }, { "abstract": "COVID-19, the disease caused by the SARS-CoV-2 virus, has been declared a pandemic by the World Health Organization, which has reported over 18 million confirmed cases as of August 5, 2020. In this review, we present an overview of recent studies using Machine Learning and, more broadly, Artificial Intelligence, to tackle many aspects of the COVID-19 crisis. We have identified applications that address challenges posed by COVID-19 at different scales, including: molecular, by identifying new or existing drugs for treatment; clinical, by supporting diagnosis and evaluating prognosis based on medical imaging and non-invasive measures; and societal, by tracking both the epidemic and the accompanying infodemic using multiple data sources. We also review datasets, tools, and resources needed to facilitate Artificial Intelligence research, and discuss strategic considerations related to the operational implementation of multidisciplinary partnerships and open science. We highlight the need for international cooperation to maximize the potential of AI in this and future pandemics.", "evidence_level": "abstract_only", "first_published": "2020-03-25", "limitations": [ "摘要未说明局限。", "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "AI应用覆盖分子（药物识别）、临床（诊断和预后）和社会（疫情和虚假信息追踪）等多个尺度。", "需要多学科合作和开放科学来最大化AI在疫情中的潜力。", "国际协作对于当下和未来疫情中发挥AI作用至关重要。" ], "method": "综述性回顾，概述了利用机器学习和人工智能应对COVID-19危机的研究，涵盖分子、临床和社会层面的应用，并回顾了数据集、工具和资源。", "metrics": { "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 5, "graph_foundation": 0.0, "graph_frontier": 0.22000000000000003, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "landscape_relevance_class": "direct", "landscape_relevance_reasons": [ "screening_direct_match" ], "topic_cluster": 5, "topic_cluster_coherence": 0.6593335309716388, "topic_cluster_relation": "direct", "topic_cluster_title": "artificial / intelligence / covid-19" }, "paper_id": "arxiv:2003.11336", "rank": 46, "relevance": "提供AI在科学（疫情应对）领域的全景式应用图谱，展示了研究分支和证据，直接支撑对AI for Science领域分支的理解。", "screening_reason": "Comprehensive review of AI applications against COVID-19, illustrating branches and evidence across scales.", "source": "arxiv", "source_url": "https://arxiv.org/abs/2003.11336v3", "summary_status": "generated", "title": "Mapping the Landscape of Artificial Intelligence Applications against COVID-19" }, { "abstract": "Deep learning bears promise for drug discovery, including advanced image analysis, prediction of molecular structure and function, and automated generation of innovative chemical entities with bespoke properties. Despite the growing number of successful prospective applications, the underlying mathematical models often remain elusive to interpretation by the human mind. There is a demand for 'explainable' deep learning methods to address the need for a new narrative of the machine language of the molecular sciences. This review summarizes the most prominent algorithmic concepts of explainable artificial intelligence, and dares a forecast of the future opportunities, potential applications, and remaining challenges.", "evidence_level": "abstract_only", "first_published": "2020-07-01", "limitations": [ "摘要未说明局限。", "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "深度学习在药物发现中具有前景，但模型通常难以被人类理解。", "需要可解释的深度学习方法以应对分子科学机器语言新叙事的需要。", "综述总结了可解释AI的突出算法概念，并展望了未来机会和挑战。" ], "method": "综述性论文，总结可解释人工智能的主要算法概念，并预测未来机遇、应用和挑战。", "metrics": { "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 2, "graph_foundation": 0.0, "graph_frontier": 0.22000000000000003, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "landscape_relevance_class": "direct", "landscape_relevance_reasons": [ "screening_direct_match" ], "topic_cluster": 2, "topic_cluster_coherence": 0.6257871472780765, "topic_cluster_relation": "direct", "topic_cluster_title": "artificial / intelligence / discovery" }, "paper_id": "arxiv:2007.00523", "rank": 47, "relevance": "展示了AI在药物发现这一分支的核心挑战（可解释性），代表了AI for Science中关于模型可理解性的重要讨论。", "screening_reason": "Review of explainable AI in drug discovery, clearly connecting a research branch to core needs and opportunities; strong evidence synthesis.", "source": "arxiv", "source_url": "https://arxiv.org/abs/2007.00523v2", "summary_status": "generated", "title": "Drug discovery with explainable artificial intelligence" }, { "abstract": "BACKGROUND: Artificial Intelligence (AI) is becoming more and more important especially in datacentric fields, such as biomedical research and biobanking. However, AI does not only offer advantages and promising benefits, but brings about also ethical risks and perils. In recent years, there has been growing interest in AI ethics, as reflected by a huge number of (scientific) literature dealing with the topic of AI ethics. The main objectives of this review are: (1) to provide an overview about important (upcoming) AI ethics regulations and international recommendations as well as available AI ethics tools and frameworks relevant to biomedical research, (2) to identify what AI ethics can learn from findings in ethics of traditional biomedical research - in particular looking at ethics in the domain of biobanking, and (3) to provide an overview about the main research questions in the field of AI ethics in biomedical research. METHODS: We adopted a modified thematic review approach focused on understanding AI ethics aspects relevant to biomedical research. For this review, four scientific literature databases at the cross-section of medical, technical, and ethics science literature were queried: PubMed, BMC Medical Ethics, IEEE Xplore, and Google Scholar. In addition, a grey literature search was conducted to identify current trends in legislation and standardization. RESULTS: More than 2,500 potentially relevant publications were retrieved through the initial search and 57 documents were included in the final review. The review found many documents describing high-level principles of AI ethics, and some publications describing approaches for making AI ethics more actionable and bridging the principles-to-practice gap. Also, some ongoing regulatory and standardization initiatives related to AI ethics were identified. It was found that ethical aspects of AI implementation in biobanks are often like those in biomedical research, for example with regards to handling big data or tackling informed consent. The review revealed current 'hot' topics in AI ethics related to biomedical research. Furthermore, several published tools and methods aiming to support practical implementation of AI ethics, as well as tools and frameworks specifically addressing complete and transparent reporting of biomedical studies involving AI are described in the review results. CONCLUSIONS: The review results provide a practically useful overview of research strands as well as regulations, guidelines, and tools regarding AI ethics in biomedical research. Furthermore, the review results show the need for an ethical-mindful and balanced approach to AI in biomedical research, and specifically reveal the need for AI ethics research focused on understanding and resolving practical problems arising from the use of AI in science and society.", "evidence_level": "abstract_only", "first_published": "2022-08-01", "limitations": [ "摘要未说明局限。", "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "AI伦理文献多描述高层原则，一些研究提出使伦理更具可操作性的方法，以弥合原则与实践之间的差距。", "生物样本库中的AI伦理问题与生物医学研究中的伦理问题相似，如大数据处理和知情同意。", "存在一些正在进行的AI伦理监管和标准化举措。" ], "method": "采用修改的主题综述方法，查询了PubMed、BMC Medical Ethics、IEEE Xplore和Google Scholar四个数据库，并补充了灰色文献检索；最终纳入57篇文献进行综述。", "metrics": { "citation_count": 39, "citation_normalized_percentile": 0.84041731, "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 4, "graph_foundation": 0.0, "graph_frontier": 0.4220535139227064, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "is_retracted": false, "landscape_relevance_class": "contextual", "landscape_relevance_reasons": [ "screening_contextual_value" ], "topic_cluster": 4, "topic_cluster_coherence": 0.6362756898177276, "topic_cluster_relation": "contextual", "topic_cluster_title": "artificial / intelligence / 2024", "venue": "Yearbook of Medical Informatics" }, "paper_id": "openalex:W4311281298", "rank": 48, "relevance": "涉及AI for Science中的伦理分支，提供了生物医学研究中AI伦理的法规、指南和工具概览，但更侧重于伦理框架而非科学结论本身。", "screening_reason": "Literature review on AI ethics in biomedical research. Relevant to AI for Science but focuses on ethical frameworks rather than core scientific conclusions or research branches. Provides useful compliance guidance but limited for understanding AI-driven discovery.", "source": "openalex", "source_url": "https://doi.org/10.1055/s-0042-1742516", "summary_status": "generated", "title": "A Literature Review on Ethics for AI in Biomedical Research and Biobanking" }, { "abstract": "We propose a hybrid Artificial Intelligence (AI) climate modeling approach that enables climate modelers in scientific discovery using a climate-targeted simulation methodology based on a novel combination of deep neural networks and mathematical methods for modeling dynamical systems. The simulations are grounded by a neuro-symbolic language that both enables question answering of what is learned by the AI methods and provides a means of explainability. We describe how this methodology can be applied to the discovery of climate tipping points and, in particular, the collapse of the Atlantic Meridional Overturning Circulation (AMOC). We show how this methodology is able to predict AMOC collapse with a high degree of accuracy using a surrogate climate model for ocean interaction. We also show preliminary results of neuro-symbolic method performance when translating between natural language questions and symbolically learned representations. Our AI methodology shows promising early results, potentially enabling faster climate tipping point related research that would otherwise be computationally infeasible.", "evidence_level": "abstract_only", "first_published": "2023-02-14", "limitations": [ "结果基于替代气候模型，而非完整气候模型。", "神经符号方法的性能仅是初步结果。", "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "该方法能够以高准确度预测AMOC（大西洋经向翻转环流）的崩溃（基于替代气候模型）。", "神经符号方法在自然语言问题与符号化学习表示之间进行翻译的初步结果展示了潜力。", "该AI方法可能加速气候临界点相关研究，否则在计算上不可行。" ], "method": "提出一种混合AI气候建模方法，结合深度神经网络和动力系统数学方法，并利用神经符号语言支持可解释性和问答。", "metrics": { "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 0, "graph_foundation": 0.0, "graph_frontier": 0.385, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "landscape_relevance_class": "contextual", "landscape_relevance_reasons": [ "screening_contextual_value" ], "topic_cluster": 0, "topic_cluster_coherence": 0.5440023663661402, "topic_cluster_relation": "direct", "topic_cluster_title": "science / scientific / intelligence" }, "paper_id": "arxiv:2302.06852", "rank": 49, "relevance": "作为AI助力科学发现（气候临界点预测）的具体实例，展示了混合AI在复杂地球系统建模中的价值，是AI for Science的一个应用分支。", "screening_reason": "Specific AI method for climate tipping points, useful as an example but not a broad synthesis; limited scope.", "source": "arxiv", "source_url": "https://arxiv.org/abs/2302.06852v1", "summary_status": "generated", "title": "Using Artificial Intelligence to aid Scientific Discovery of Climate Tipping Points" }, { "abstract": "Recent advances in large language models (LLMs) have enabled a new class of AI agents that automate multiple stages of the data science workflow by integrating planning, tool use, and multimodal reasoning across text, code, tables, and visuals. This survey presents the first comprehensive, lifecycle-aligned taxonomy of data science agents, systematically analyzing and mapping forty-five systems onto the six stages of the end-to-end data science process: business understanding and data acquisition, exploratory analysis and visualization, feature engineering, model building and selection, interpretation and explanation, and deployment and monitoring. In addition to lifecycle coverage, we annotate each agent along five cross-cutting design dimensions: reasoning and planning style, modality integration, tool orchestration depth, learning and alignment methods, and trust, safety, and governance mechanisms. Beyond classification, we provide a critical synthesis of agent capabilities, highlight strengths and limitations at each stage, and review emerging benchmarks and evaluation practices. Our analysis identifies three key trends: most systems emphasize exploratory analysis, visualization, and modeling while neglecting business understanding, deployment, and monitoring; multimodal reasoning and tool orchestration remain unresolved challenges; and over 90% lack explicit trust and safety mechanisms. We conclude by outlining open challenges in alignment stability, explainability, governance, and robust evaluation frameworks, and propose future research directions to guide the development of robust, trustworthy, low-latency, transparent, and broadly accessible data science agents.", "evidence_level": "abstract_only", "first_published": "2025-10-05", "limitations": [ "摘要未说明局限，但分析的覆盖范围限于所映射的45个系统。", "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "大多数系统强调探索性分析、可视化和建模，而忽视业务理解、部署和监控。", "多模态推理和工具编排仍是未解决的挑战。", "超过90%的系统缺乏明确的信任和安全机制。" ], "method": "对45个基于LLM的数据科学代理系统进行综述，提出一种与生命周期一致的分类法，将系统映射到端到端数据科学过程的六个阶段，并沿五个设计维度进行注释和分析。", "metrics": { "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 0, "graph_foundation": 0.0, "graph_frontier": 0.49500000000000005, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "landscape_relevance_class": "contextual", "landscape_relevance_reasons": [ "screening_contextual_value" ], "topic_cluster": 0, "topic_cluster_coherence": 0.5440023663661402, "topic_cluster_relation": "direct", "topic_cluster_title": "science / scientific / intelligence" }, "paper_id": "arxiv:2510.04023", "rank": 50, "relevance": "探讨了AI在自动化数据科学工作流中的应用，这与AI for Science中数据密集型和模型构建环节相关，但更偏向通用数据科学而非特定科学发现。", "screening_reason": "Survey of LLM data science agents, related to AI for Science workflows but not directly addressing scientific discovery conclusions.", "source": "arxiv", "source_url": "https://arxiv.org/abs/2510.04023v1", "summary_status": "generated", "title": "LLM-Based Data Science Agents: A Survey of Capabilities, Challenges, and Future Directions" }, { "abstract": "The CONSORT 2010 (Consolidated Standards of Reporting Trials) statement provides minimum guidelines for reporting randomised trials. Its widespread use has been instrumental in ensuring transparency when evaluating new interventions. More recently, there has been a growing recognition that interventions involving artificial intelligence (AI) need to undergo rigorous, prospective evaluation to demonstrate impact on health outcomes.The CONSORT-AI extension is a new reporting guideline for clinical trials evaluating interventions with an AI component. It was developed in parallel with its companion statement for clinical trial protocols: SPIRIT-AI. Both guidelines were developed through a staged consensus process, involving a literature review and expert consultation to generate 29 candidate items, which were assessed by an international multi-stakeholder group in a two-stage Delphi survey (103 stakeholders), agreed on in a two-day consensus meeting (31 stakeholders) and refined through a checklist pilot (34 participants).The CONSORT-AI extension includes 14 new items, which were considered sufficiently important for AI interventions, that they should be routinely reported in addition to the core CONSORT 2010 items. CONSORT-AI recommends that investigators provide clear descriptions of the AI intervention, including instructions and skills required for use, the setting in which the AI intervention is integrated, the handling of inputs and outputs of the AI intervention, the human-AI interaction and providing analysis of error cases.CONSORT-AI will help promote transparency and completeness in reporting clinical trials for AI interventions. It will assist editors and peer-reviewers, as well as the general readership, to understand, interpret and critically appraise the quality of clinical trial design and risk of bias in the reported outcomes.", "evidence_level": "abstract_only", "first_published": "2020-09-09", "limitations": [ "摘要未说明局限。", "仅依据公开摘要，未核验全文方法、实验设置与结论边界。" ], "main_findings": [ "CONSORT-AI扩展包含14个新增项目，认为这些项目对AI干预措施足够重要，应常规报告。", "建议提供AI干预的清晰描述，包括使用说明和所需技能、集成环境、输入输出处理、人机交互以及错误案例分析。", "该指南有助于提高AI干预临床试验报告的透明度和完整性，帮助编辑、同行评审员和读者评估试验设计的质量和偏倚风险。" ], "method": "采用阶段性共识流程开发报告指南，包括文献回顾和专家咨询以产生29个候选项目，随后进行两轮德尔菲调查（103名利益相关者）、共识会议（31人）和试点测试（34名参与者），最终形成CONSORT-AI扩展。", "metrics": { "citation_count": 632, "citation_normalized_percentile": 0.99254658, "graph_bridge": 0.0, "graph_citation_in_degree": 0, "graph_community": 4, "graph_foundation": 0.0, "graph_frontier": 0.35604208461359366, "graph_reference_out_degree": 0, "graph_structural_role": "frontier", "graph_topology_community": 0, "is_retracted": false, "landscape_relevance_class": "contextual", "landscape_relevance_reasons": [ "screening_contextual_value" ], "topic_cluster": 4, "topic_cluster_coherence": 0.6362756898177276, "topic_cluster_relation": "contextual", "topic_cluster_title": "artificial / intelligence / 2024", "venue": "BMJ" }, "paper_id": "openalex:W3083992516", "rank": 52, "relevance": "为AI在医疗健康领域的临床试验提供标准化报告指南，是AI for Science中确保科学质量与透明性的关键支撑，但属于临床评估分支。", "screening_reason": "Rigorous reporting guideline for AI clinical trials; relevant to AI in healthcare but not broad AI for Science consensus.", "source": "openalex", "source_url": "https://doi.org/10.1136/bmj.m3164", "summary_status": "generated", "title": "Reporting guidelines for clinical trial reports for interventions involving artificial intelligence: the CONSORT-AI Extension" } ], "tags": [ "AI for Science", "领域综述", "核心共识", "数据基础设施", "物理信息神经网络", "科学大语言模型", "自主实验", "实验验证" ], "title": "AI for Science：从算法突破到科研基础设施——四条线索下的可靠共识与待验证推断", "updatedAt": "2026-08-13" }