单向量检索把整段文本压成一个点，cross-encoder 准但每对都要过一次模型。ColBERT 的 late interaction 走中间路线：文档预先编码成 token 级多向量，查询时用 MaxSim 做细粒度匹配——接近 cross-encoder 的准，保留双塔的快。本文讲清它准在哪、贵在哪、该放在检索流水线的哪一层。

Learning general-purpose robot policies requires large-scale real-world interaction data, yet collecting robot demonstra…

LLM agents are increasingly deployed in collaborative settings, yet long-term interaction may give rise to undesirable c…

Interaction understanding in 3D scenes requires a joint description of movable parts, their motion, and the regions thro…

Natural interaction in digital and physical environments requires continuous perception and timely responses. Spoken dia…

Late-interaction retrieval is the state-of-the-art for visual document search, but it pays for its accuracy in storage. …

In this work, we present Gander, an end-to-end model that unifies omni perception, realtime interaction, and agentic cap…

Generating realistic human-object interactions (HOI) in complex 3D scenes requires two complementary capabilities: reaso…

Egocentric world models predict first-person observations conditioned on an agent's actions, but most focus on a single …

Streaming video LLMs must retain evidence before its relevance to future tasks is known and respond when sufficient evid…

Spatial reasoning is essential for vision-language models (VLMs) to understand and act in the physical world. Reasoning …

Language model agents increasingly operate over long interaction streams, requiring online adaptation to distribution sh…

Large-scale egocentric human interaction data is becoming an important source of physical supervision for embodied learn…

GPU-parallel simulation provides abundant robot interaction, but existing benchmarks rarely combine this scale with hete…

Oct 1, 2026 Tavus has introduced Griffin, what the company calls the first "Human Interaction Model" (HIM),** a class of…

Long-context LLM agents accumulate interaction histories that strain KV-cache memory and attention computation. Although…

Tool-calling agents have become central to enterprise AI, yet training and evaluating them at scale remains severely con…

Agent harnesses, the external systems that mediate model-environment interaction, can substantially improve agent perfor…

Human video datasets offer an abundant and diverse source of interaction data that can complement expensive real-robot d…

Computer-use agents (CUAs) have advanced along two separate lines: graphical interaction and software development throug…

World modeling enables intelligence to anticipate consequences, guide interventions, and learn from interaction. Yet pre…

In collaborative tasks with asymmetric information, participants coordinate their understanding through interaction. We …

Realtime spoken interaction demands deep reasoning, prompt responses, and fluid turn-taking. We present StepAudio 3 Real…

双语育儿被焦虑裹挟得最深：错过「语言敏感期」就晚了？双语会让孩子说话更晚？外教和启蒙动画哪个管用？语言学与发展心理学的答案比营销号冷静得多——双语不会搞混孩子，敏感期的窗口远比传闻宽，而家庭里最重要的变量不是机构、不是教材，是语言「活着」。

!(https://www.gstatic.com/images/branding/googleg/1x/googleg_standard_color_128dp.png)Add as a preferredsource on Google…

At a glance Brings rich interactivity to AI-generated charts: Flint-interactive enables agents to create data visualizat…

Governments need modern solutions to serve the public while keeping sensitive information under their control. Open mode…

To increase productivity, accelerate innovation and create new opportunities for economic growth, Europe’s industrial le…

NVIDIA cuDSS, launched today in general availability, is an NVIDIA CUDA-X library that shifts a critical part of enginee…

General availability of the instances is coming next month, with planned expansion of availability to Europe — bringing …

Robots must act under physical and social constraints while adapting to changes in their environment and human needs. Ex…

Despite substantial progress in Deep Search agents, their ability to conduct end-to-end research in real-world open Web …

Effective test-time personalization requires LLM agents to adapt to query-dependent user preferences from lightweight in…

!(https://www.gstatic.com/images/branding/googleg/1x/googleg_standard_color_128dp.png)Add as a preferredsource on Google…

!(https://www.gstatic.com/images/branding/googleg/1x/googleg_standard_color_128dp.png)Add as a preferredsource on Google…

论文提出涵盖十个领域的 τ-Emotion-Bench 基准，在受控的情绪化用户交互下评估工具调用智能体，并配套 Tracer 数据生成管线用于构建任务与采集训练轨迹。

论文提出StreamErase，通过将双向基础模型的擦除能力迁移到单向流式生成器，在单张GPU上实现最高26 FPS、亚秒级延迟的流式视频物体移除。

论文提出“战略选择”元博弈框架：在博弈收益矩阵未知且需在线学习的情况下，从模型组合中动态选取LLM参与博弈，弥补现有路由器不考虑对手策略自适应的不足。

Apple revealed in a regulatory filing that it has reached an agreement to bring on team members and technology from pers…

In a tan polo shirt, brown leather belt, and round bolo tie, MIT PhD student Wil Norton kind of looks like a park ranger…

Rather than downloading another app, a growing number of agents can simply be texted like an ordinary person. You text i…

Ever since people first seriously contemplated giving machines an intelligence modeled on our own, there has never been …

Computer-use agents can perform increasingly complex GUI tasks, yet unfamiliar software and workflows requiring human do…

Turning a simulation idea into a working application means assembling assets, connecting physics and rendering, and chec…

文章介绍 SWEeper-Bench 基准，包含来自 200 个广泛使用的开源 Web 应用真实修复的 200 个任务，用于评估智能体在多步交互场景下自主发现并修复缺陷的能力，而现有基准多假设缺陷已被识别。

At a glance Harnessed Agentic RL: Microsoft Research Asia introduces a training paradigm in which the same agent harness…

Large foundation models have accelerated progress toward general-purpose agents that interact with humans and other agen…

Computer-use agents (CUAs) are rapidly improving, yet synthesizing diverse and reliable software environments at scale r…

Learning to act in unfamiliar environments requires agents to infer how the world works and revise that understanding as…

When retrieved evidence contradicts an agent's prior beliefs, does it revise its answer, acknowledge uncertainty, or per…