Curating the frontier of Artificial Intelligence

Agent Memory: Characterization and System Implications of Stateful Long-Horizon Workloads

LLM 智能体越来越多地部署在需要对长期交互历史进行持续推理的长期任务上。大规模地意识到这一点......

LLM agents are increasingly deployed on long-horizon tasks requiring sustained reasoning over extended interaction histories. Realizing this at scale ...

Vortex: Efficient and Programmable Sparse Attention Serving for AI Agents

随着生成长度的不断增长，稀疏注意力对于服务大型语言模型 (LLMs) 变得越来越重要。然而, 部署...

Sparse attention is becoming increasingly important for serving large language models (LLMs) as generation lengths continue to grow. However, deployin...

Benchmark Everything Everywhere All at Once

基准通过提供标准化和明确的绩效衡量标准，是评估和推进 LLM 和 MLLM 的基础。然而,他们的...

Benchmarks are fundamental for evaluating and advancing LLMs and MLLMs by providing standardized and explicit measures of performance. However, their ...

Goedel-Architect: Streamlining Formal Theorem Proving with Blueprint Generation and Refinement

我们引入了 Goedel-Architect, ，这是一个在 Lean 4 中以蓝图生成和细化为中心的形式定理证明的代理框架。一张蓝图...

We introduce Goedel-Architect, an agentic framework for formal theorem proving in Lean 4 centered on blueprint generation and refinement. A blueprint ...

MLEvolve: A Self-Evolving Framework for Automated Machine Learning Algorithm Discovery

大型语言模型 (LLM) 代理越来越多地应用于长期任务，例如科学发现和机器学习工程 (MLE), ...

Large language model (LLM) agents are increasingly applied to long-horizon tasks such as scientific discovery and machine learning engineering (MLE), ...

Beneath Nvidia AI chips, Chinese PCBs raise security concerns in U.S.

印刷电路板几乎位于每个芯片的下面,，几乎是每个电子产品的必需品。它们构成了蓬勃发展的安静但至关重要的一部分......

Printed circuit boards sit underneath almost every chip, a necessity in nearly every electronic. They make up a quiet but crucial piece of the booming...

Knowledge Index of Noah's Ark

LLM 的知识基准面临三个问题: 扩展驱动的设计无法实现学科代表性; 固定付款注释...

Knowledge benchmarks for LLMs face three issues: scaling-driven designs that do not operationalize disciplinary representativeness; flat-payment annot...

MIT researchers teach AI models to interpret charts

为了在快节奏的,全球市场,中加速和完善决策，企业可以部署生成人工智能模型来帮助...

To accelerate and refine decision-making in a fast-paced, global marketplace, enterprises may deploy generative artificial intelligence models to help...

NSF renews support for MIT

麻省理工学院领导的人工智能和基本交互研究所 (IAIFI) 获得了国家科学基金会的重新支持...

The MIT-led Institute for Artificial Intelligence and Fundamental Interactions (IAIFI) has received renewed support from the National Science Foundati...

PATH to boost AI training and career opportunities for industry

MIT, 与佐治亚州立大学和不断增长的教育机构网络, 合作，宣布在 PATH (Pathway 下扩大工作...

MIT, in collaboration with Georgia State University and a growing network of educational institutions, has announced expanded work under PATH (Pathway...