Curating the frontier of Artificial Intelligence

Tod Machover receives George Peabody Medal for contributions to music and technology

作为作曲家和音乐技术先驱, Machover 通过他参与的工作帮助艺术家和观众扩大了音乐的可能性...

As a composer and music tech pioneer, Machover has helped expand music’s possibilities for artists and audiences alike through his work in participato...

Teaching AI agents to ask better questions by playing “Battleship”

CSAIL 和 SEAS 学者通过围绕提问和回答自然语言问题重新构建游戏，增加了一些变化。在他们的 “ 协作战舰中...

CSAIL and SEAS scholars added a twist by reframing the game around asking and answering natural language questions. In their “Collaborative Battleship...

SafeSteer: Localized On-Policy Distillation for Efficient Safety Alignment

将大型语言模型 (LLMs) 与人类价值观保持一致通常会降低其一般能力,，称为对齐税。现有方法减轻了...

Aligning Large Language Models (LLMs) with human values often degrades their general capabilities, termed the alignment tax. Existing methods mitigate...

Tracking the Behavioral Trajectories of Adapting Agents

诸如技能文件, 内存文件, 和行为配置文件之类的文本文件在定义现代代理的行为方式方面发挥着核心作用。通过编辑...

Text files such as skill files, memory files, and behavioral configuration files play a central role in defining how modern agents act. Through edits ...

ClinEnv: An Interactive Multi-Stage Long Horizon EHR Environment for Agents

临床实践不是从列举的选项中选择答案: 医生逐渐收集异构信息并致力于...

Clinical practice is not the selection of an answer from enumerated options: a physician gathers heterogeneous information incrementally and commits t...

AutoSci: A Memory-Centric Agentic System for the Full Scientific Research Lifecycle

科学研究传统上是人力密集型,，要求研究人员协调文献, 想法, 实验, 手稿, 和评论...

Scientific research has traditionally been human-intensive, requiring researchers to coordinate literature, ideas, experiments, manuscripts, and revie...

LinTree: Improving LLM Reasoning with Explicitly Structured Search Histories

大型语言模型 (LLMs) 通常通过生成探索和修改部分解决方案的中间轨迹来解决推理问题。从搜索...

Large language models (LLMs) often solve reasoning problems by generating intermediate traces that explore and revise partial solutions. From a search...

Choosing the Lens: Strategic Perspective Activation in Context-Dependent Argumentation

相同的论点常常需要在不同的外部制度下进行评估。对政权有影响力的特工拥有战略杠杆，可以...

The same arguments often need to be evaluated under different external regimes. An agent with influence over the regime has a strategic lever that sta...

ProjectionBench: Evaluating Scientific Hypothesis Generation in LLMs Under Progressive Information Disclosure

科学发现本质上是一个创造性和不确定性的过程,，需要超出已知知识回忆的推理。虽然许多基准...

Scientific discovery is an inherently creative and uncertain process, requiring reasoning beyond the recall of known knowledge. While many benchmarks ...

MIRA: Mid-training Rubric Anchoring for Source-Aware Data Selection

中期培训已成为现代 LLM 发展, 的一个重要阶段，在最终的后期培训之前，使用大规模策划的混合物来增强能力。

Mid-training has become an important stage in modern LLM development, using large-scale curated mixtures to strengthen capabilities before final post-...