研究表明，机器学习在公共部门的成功，将更少依赖模型准确率的突破，而更多依赖机构能否构建出透明、可复现、可问责且受公民信任的数据基础设施。It is shown that the success of machine learning in the public sector will depend less on breakthroughs in model accuracy and more on the ability of institutions to engineer transparent, reproducible, and accountable data infrastructures that citizens can trust.
论文
9 张论文卡片 · 工程化 · OA 绿色
条目 G: 公共部门 ML Pipeline 工程教训（含性能数据表）
条目 F: Google 企业定制 LLM — 代码转换实战数据
介绍 Gemini for Google (GfG)，一款面向 Google 内部软件工程生态的 Gemini 专用适配版本，涵盖从构建万亿 token 的专有数据集到采用可缓解灾难性遗忘的中段训练策略的完整过程。G Gemini for Google (GfG)}, an adaptation of Gemini specialized for Google's internal software engineering ecosystem, is introduced, from curating a trillion-token proprietary dataset to implementing a mid-training strategy that mitigates catastrophic forgetting.
条目 E-NF2：MLOps 架构指南 — 25 条模型集成/部署规范（灰色文献综述）
本文贡献了一组在架构层面具有重要意义的 MLOps 集成与部署指南（共 25 条），分为五类，并阐述其对整体系统架构的影响。This work contributes a collection of 25 architecturally significant MLOps guidelines for model integration and deployment, organized into five categories, and describes their impact on the overall system architecture.
⑥ "How are MLOps Frameworks Used in Open Source Projects"（arXiv:2601.18591）
⑥ "How are MLOps Frameworks Used in Open Source Projects"（arXiv:2601.18591）
对八款主流开源 MLOps 框架的实践使用与功能增强需求进行调查，结果显示 MLOps 框架很少被直接开箱即用，也较少集成进 GitHub Workflows，开发者更多通过其 API 在项目中实现自定义功能。Investigating the practical use and desired feature enhancements of eight popular open-source MLOps frameworks indicates that users mainly ask for enhancements to core features of the frameworks, but also better API exposure and CI/CD integration.
⑤ MLOps系统综述（arXiv:2604.16371）
⑤ MLOps系统综述（arXiv:2604.16371）
对聚焦MLOps工具的学术文献进行系统综述，揭示其功能、范围及其旨在解决的挑战，并突出真实MLOps pipeline中各工具间互操作性的重要性。A systematic review of the academic literature focused on MLOps tools is conducted to reveal their function, scope, and the challenges they are designed to address and highlight the importance of interoperability across MLOps tools in real-world MLOps pipelines.
3️⃣ arXiv · MatryoshkaLoRA（⭐⭐⭐⭐ 值得关注）
3️⃣ arXiv · MatryoshkaLoRA（⭐⭐⭐⭐ 值得关注）
本文提出 MatryoshkaLoRA，一种受 Matryoshka 启发、面向 LoRA 的通用训练框架，通过在已有 LoRA adapter 之间插入一个固定的、经精心设计的对角矩阵来按比例缩放其子秩，从而学习到准确的层次化低秩表示。MatryoshkaLoRA is proposed, a general, Matryoshka-inspired training framework for LoRA that learns accurate hierarchical low-rank representations by inserting a fixed, carefully crafted diagonal matrix between the existing LoRA adapters to scale their sub-ranks accordingly.
3. Kubernetes for GenAI Inference（arXiv:2602.04900v2）
Kubernetes for GenAI Inference（arXiv:2602.04900v2）
这些结果表明，Kueue、DAS 与 GAIE 等互补组件构成了一个高性能的协同平台，证明了 Kubernetes 能够作为承载高要求 GenAI 工作负载的统一底座。These findings illustrate that these complementary components (Kueue, DAS, and GAIE) form a cohesive, high-performance platform, proving Kubernetes' capability to serve as a unified foundation for demanding GenAI workloads.
A Comprehensive Survey on Transfer Learning
迁移学习全面综述
本综述尝试连接并系统化现有迁移学习研究，全面总结与阐释迁移学习的机制与策略，帮助读者更好地理解当前研究现状与思路。This survey attempts to connect and systematize the existing transfer learning research studies, as well as to summarize and interpret the mechanisms and the strategies of transfer learning in a comprehensive way, which may help readers have a better understanding of the current research status and ideas.
2010.06047
2010.06047
总结了利用人工智能以及语音与语言处理预测阿尔茨海默病相关认知衰退的现有研究发现。The existing findings on the use of artificial intelligence, speech, and language processing to predict cognitive decline in the context of Alzheimer’s disease are summarized.