Curating the frontier of Artificial Intelligence

Agentic Context Management: Solving Agent Memory and Cost by Treating Them as Lifecycle and Architecture Problems

生产型 AI 代理的故障较少是由于无法良好推理而导致的，而更常见的是因为他们无法管理推理中的内容......

Production AI agents' failures are less often due to an inability to reason well and more often because they cannot manage what is in their reasoning ...

Same Dangerous Objective, Opposite Advice: Direct Exposure versus Multi-Agent Mediation

即使是当前的高能力法学硕士，在直接显示危险目标时也会比其他代理转变并传递其方向时显得更安全......

Even a current high-capability LLM can appear safer when shown a dangerous objective directly than when other agents transform and relay its direction...

The Boundaries of Automation: A Theory of Persistent Human Participation

人工智能的快速进步强化了长期以来对自动化:的追求，尽可能用算法代替人类参与。进出口...

The rapid progress of AI has intensified the long-standing pursuit of automation: replacing human participation with algorithms wherever possible. Imp...

MIRROR: Learning from the Other View for Multi-Modal Reasoning

与表现出强大推理能力的大型语言模型 (LLMs) 不同, 视觉语言模型 (VLMs) 即使在...上也很难进行视觉推理,

Unlike large language models (LLMs) that exhibit strong reasoning capabilities, vision-language models (VLMs) struggle with visual reasoning, even on ...

CXMT is sparking fears of a cash drain before blockbuster IPO

长鑫存储科技'大规模上市引发了人们的担忧，担心其上市可能会从中国股市吸走现金,，因为投资者筹集资金...

ChangXin Memory Technologies' massive listing is stoking fears that its market debut could pull cash from Chinese equities, as investors raise funds t...

OpenForgeRL: Train Harness-native Agents in Any Environment

现代 AI 代理依靠复杂的推理工具（例如 Claude Code, Codex, 和 OpenClaw）来驱动多轮推理, 工具使用, 并访问...

Modern AI agents rely on elaborate inference harnesses such as Claude Code, Codex, and OpenClaw to drive multi-turn reasoning, tool use, and access to...

Beyond Sycophancy: Structured Resistance and Compliance in LLM Moral Reasoning

构建社会校准的大型语言模型,，它可以向他人学习，而不是简单地屈服于他们, 需要的不仅仅是减少阿谀奉承......

Building socially calibrated large language models, which can learn from others without simply yielding to them, requires more than reducing sycophanc...

U.S., other nations back open

成都, 中国— 随着公司发布更强大的开源模型, 一种新的多元...，政府越来越希望控制人工智能,

CHENGDU, China — Governments increasingly want to control artificial intelligence, as companies release more powerful open-source models, a new multil...

Unsupervised Consensus-Based Anomaly Detection for Spatiotemporal Malaria Incidence in Ghana

将共识异常检测框架应用于加纳(2014-2023) 的每月疟疾监测数据，以识别非典型传播模式...

A consensus anomaly detection framework was applied to monthly malaria surveillance data from Ghana (2014-2023) to identify atypical transmission patt...

Working to automate nuclear plant operations

要使核能被视为可行的清洁能源,，它的价格必须具有竞争力且生产经济。本科毕业后...

For nuclear to be considered as a viable clean energy source, it has to be competitively priced and economical to produce. After an undergraduate degr...