TechFeed

我们如何通过人工与 LLM 辅助标注相结合来训练 Dash 的搜索排序模型。

How we train Dash's search ranking models with a mix of human and LLM-assisted labeling.

随着 AI 模型规模扩大，它们的洞察力反而下降，而非提升。为确保模型持续学习，我们需要缩短推理时间。

As AI models grow larger, they become less insightful, not more. To ensure that they continue to learn, we need to reduce their inference time.

机器感知

Machine Perception

亚马逊学者 Aravind Srinivasan 合著了2014年关于拉美民事动荡预测的论文，该论文在2025年 KDD 获得了历时考验奖.

Amazon Scholar Aravind Srinivasan coauthored a 2014 paper about forecasting civil unrest in Latin America, which won a test-of-time award at KDD 2025.

让 Dropbox Dash 等产品对个人和企业可用意味着要应对效率和资源使用方面的新挑战。

Making products like Dropbox Dash accessible to individuals and businesses means tackling new challenges around efficiency and resource use.

从 Claude Code 到 Cursor，我们在 Dropbox 大量采用 AI 编码工具。早期结果令人鼓舞，但仍有许多未解之问，关于如何最有效地使用这些工具以及它们能产生最大影响的场景。推动这场讨论…

From Claude Code to Cursor, we're big adopters of AI coding tools at Dropbox. The early results have been promising, but there are still a lot of open questions about how to work with these tools most effectively and where they can have the most impact. To push this conversation…

算法与理论

Algorithms & Theory

人机交互与可视化

Human-Computer Interaction and Visualization

LongCat-Flash-Lite是一款拥有 685 亿参数，每次推理仅激活 29 亿~ 45 亿参数的轻量化 MoE 模型。通过将超过 300 亿参数高效用于嵌入层，LongCat-Flash-Lite 不仅超越了参数量等效的 MoE 基线模型，还在与同规模现有模型的对比中展现出卓越的竞争力，尤其在智能体与代码领域表现突出。

LongCat-Flash-Lite is a lightweight MoE model with 68.5 billion parameters, activating only 2.9-4.5 billion parameters per inference. By efficiently using more than 30 billion parameters in the embedding layer, LongCat-Flash-Lite not only outperforms MoE baseline models with equivalent parameter counts, but also shows superior competitiveness against same-scale existing models, particularly in the agent and code fields.

气候与可持续发展

Climate & Sustainability

教育创新

Education Innovation

亚马逊科学家与斯坦福研究人员的早期会面促成了 cvc5，这一开源工具如今每天在 AWS 上支持约十亿次自动推理检查。

An early meeting between Amazon scientists and Stanford researchers led to cvc5, an open-source tool now powering approximately one billion automated-reasoning checks across AWS every day.

在学术竞赛中首次，学生可以定制前沿 AI 模型以构建可信的软件代理

For the first time in an academic competition, students can customize frontier AI models to build trusted software agents

推动 AI 发展不仅需要突破性的模型，还依赖于进行实验、验证假设并分享学习成果的构建者和研究者社区。这一信念指引着亚马逊围绕 Amazon Nova——亚马逊的 AI 产品组合——与开发者和学术界的互动方式…

Advancing AI requires more than breakthrough models. It depends on communities of builders and researchers who experiment, test assumptions, and share what they learn. That belief is guiding how Amazon engages developers and academics around Amazon Nova, Amazon’s portfolio of AI…

每场 NFL 比赛都会从 22 名装有 RFID 的球员产生数百万个追踪数据点。75 个在 AWS 上运行的机器学习模型在不到一秒的时间内处理这些数据，将足球转变为每一次动作都被测量、建模并即时分析的运动。

Every NFL game generates millions of tracking data points from 22 RFID-equipped players. Seventy-five machine learning models running on AWS process that data in under a second, transforming football into a sport where every movement is measured, modeled, and instantly analyzed.