Curating the frontier of Artificial Intelligence

CaMBRAIN: Real-time, Continuous EEG Inference with Causal State Space Models

脑电图(EEG) 是监测脑电活动的关键, 非侵入性方法。脑电图的跨度可以从几秒钟到......

Electroencephalography (EEG) is a critical, non-invasive method to monitor electrical brain activity. EEGs can span anywhere from a couple seconds to ...

Calibrating Conservatism for Scalable Oversight

能够自主规划和扩展环境交互的代理人工智能系统提出了一个基本的控制问题:人类如何维持...

Agentic AI systems capable of autonomous planning and extended environmental interaction pose a fundamental control problem: how can humans maintain m...

Media Advisory: MIT to establish regional quantum hub

QSL 将位于麻省理工学院校园 39 号楼，将作为一个拥有现代化实验基础设施的多学科量子中心。乙...

The QSL will be located at Building 39 on the MIT campus and will serve as a multi-disciplinary quantum hub with modern experimental infrastructure. B...

Natural Language Query to Configuration for Retrieval Agents

现代检索代理公开了许多配置选择 - LLM, 检索器, 文档数, 跳数, 和合成策略 - 每个形状...

Modern retrieval agents expose many configuration choices -- LLM, retriever, number of documents, number of hops, and synthesis strategy -- each shapi...

MUSE-Autoskill: Self-Evolving Agents via Skill Creation, Memory, Management, and Evaluation

大型语言模型 (LLM) 代理依靠可重用技能来解决复杂任务,，但现有的技能创建方法通常将技能视为孤立的......

Large language model (LLM) agents rely on reusable skills to solve complex tasks, but existing skill creation approaches often treat skills as isolate...

Claw-Anything: Benchmarking Always-On Personal Assistants with Broader Access to User's Digital World

大型语言模型代理越来越多地被设想为永远在线的个人助理，可以访问用户 27 年代数字世界中的任何相关内容......

Large language model agents are increasingly envisioned as always-on personal assistants with access to anything relevant in the user's digital world....

From Model Scaling to System Scaling: Scaling the Harness in Agentic AI

本文研究了代理 AI 的下一个主要瓶颈，因为系统扩展, 不仅是模型扩展:，还有可审计, 持久, 模块化, 的设计以及...

This paper studies the next major bottleneck in agentic AI as system scaling, not only model scaling: the design of auditable, persistent, modular, an...

MobileGym: A Verifiable and Highly Parallel Simulation Platform for Mobile GUI Agent Research

我们推出 MobileGym, 一个浏览器托管的, 轻量级, 完全可控的环境，适合日常移动使用, 目标是交互保真度，无需重新...

We present MobileGym, a browser-hosted, lightweight, fully controllable environment for everyday mobile use, targeting interaction fidelity without re...

From Raw Experience to Skill Consumption: A Systematic Study of Model-Generated Agent Skills

语言代理通过重用\emph{skills}（从过去的经验中提取的结构化程序工件）不断改进。特别是, \emph{...

Language agents increasingly improve by reusing \emph{skills} -- structured procedural artifacts distilled from past experience. In particular, \emph{...

SkillOpt: Executive Strategy for Self-Evolving Agent Skills

如今的代理技能是手工制作的, 一次性生成的, 或通过松散控制的自我修订, 演变而来的,，其中没有一个表现得像深度学习...

Agent skills today are hand-crafted, generated one-shot, or evolved through loosely controlled self-revision, none of which behaves like a deep-learni...