Dropbox reduced its backend monorepo from 87GB to 20GB by optimizing Git delta compression in collaboration with GitHub. The changes improved clone times, CI performance, and developer velocity,...

AI poses major challenges for green IT: each query consumes vast energy, GPU chips last only 2-3 years, and costs stay hidden from users. Regulatory frameworks like the EU AI Act fall short on...

In today’s fast-paced financial environment, ensuring that core systems are secure, high-performing, and cost-effective is essential for maintaining a competitive edge. For Insignia Financial,...

本研究提出了Baguan模型，通过战略预训练减少天气预测中的过拟合问题。该模型基于自我监督的Siamese自编码器，实验结果表明其在中期天气预测中优于传统方法。

本研究提出了ParaPO后训练方法，旨在减少语言模型在非敌对环境中对预训练数据的逐字复制问题。该方法通过优化模型偏好改写版本，显著降低了无意复制现象，同时保持了模型的整体效用。

Jorma Rauhala, president and CEO of Kesko Corporation, reveals the success factors of Kesko’s data and AI journey.

本研究探讨了在工程教育中整合AI工具（如ChatGPT和GitHub Copilot）以提高设计和制造效率。通过这些工具，学生能够快速原型化复杂项目，提升成果复杂性并增强决策支持。

Reducing emissions in Tier 2 production is one of the fashion industry’s most complex challenges. Six actions can help make industry players more sustainable.

Discover how MySQL Enterprise Edition helps government agencies reduce IT costs by up to 90% while enhancing security, scalability, and efficiency. Learn how transitioning from costly legacy...

该研究提出了一种基于学习的搜索空间减少方法，解决了现有神经组合优化方法在大规模问题上的推广能力不足。通过动态选择有前景的候选节点，显著降低搜索空间，同时保持解决方案质量。实验表明，该方法在100节点实例上训练后，成功推广至100万节点的旅行商问题及其他车辆路径问题。

本研究提出NOVA框架，通过内部一致性探测和语义等效识别，筛选高质量数据，解决大语言模型在指令调优阶段因陌生知识导致的幻觉问题，从而提升模型遵循指令的能力。