SPADE: Self-Play in Adaptive Synthetic Executable Environments· SPADE: 自适应合成环境中的自玩
Continuous self-improvement requires an ever-expanding pool of self-generated, diverse, ad…
Continuous self-improvement requires an ever-expanding pool of self-generated, diverse, ad…
ChildSafeAds is a shared task on commercial content in YouTube videos likely to reach chil…
Language-model agents can communicate through continuous hidden states that are invisible …
Frontier language models are compared, marketed, and benchmarked on capability -- what the…
We present a novel application of embedding-based dynamic topic modeling techniques to det…
When an expert corrects an LLM assistant's error, the correction usually dies with the ses…
An agent still learning its environment should be cautious while ignorant and bold once co…
Large language model (LLM) agents can now post-train an LLM end-to-end. They can write cod…
We present Eureka, a task-conditioned Meta-Agent architecture that compiles long-horizon t…
Accurate and responsible medical question answering (QA) is important in healthcare, where…
Released in 2025, Institutional Books: Harvard Library (IB-HL) is a collection of 983,004 …
Accurately extracting nuanced, contextualized data from research articles is laborious and…
Large language models (LLMs) are increasingly paired with verifiers (step checkers, self-c…
Debates have recently emerged as a useful methodology for agentic AI to improve performanc…
Vision Language Models (VLMs) have recently emerged as a critical component of multimodal …
As AI systems are increasingly deployed in safety-critical application domains (e.g., auto…
Reinforcement-learning training of reasoning LLMs (e.g., GRPO) is expensive and requires a…
Oral diseases affect billions of people worldwide, underscoring a pressing need for accura…
Agent frameworks increasingly package procedural knowledge as skills: instruction files an…
Last-mile delivery aims to handle dynamically arriving orders with couriers while modeling…
Utilities lose a substantial share of treated water to leakage, yet rarely trust artificia…
Arguments often omit premises or claims, forming enthymemes. We study pairwise logical sel…
Agents improve quickly against a reliable automatic metric and stall without one, and the …
This paper proposes a multi-agent framework built on CrewAI [1] for conversational busines…
Emotion Ball 是一套面向 AI 助手的表情引擎:32 种状态表情全部由纯 SVG 与原生 JavaScript 实时驱动,零框架、零图片资源。AI 侧只需输出一个 em…
19 款可独立安装的 AI 视觉风格 Skill:挑一款、上传一张照片,在 Codex 里直接生成成品。作者 @威比 Hunter Wei.(抖音、小红书同名)
载重4.2吨、容积19.32m³
8月21日,摩根大通在最新研究报告中指出,阿里云当前12%利润率被系统性低估,成熟态ROIC应接近20%。原因在于,过去几个季度Capex投放快,大量GPU和数据中心资产刚投产、仍…
明略科技(2718.HK)与海康机器人联合参展2026WRC,聚焦商业服务领域展示具身智能落地进展。
8月21日,全球领先的消费级AR品牌雷鸟创新(RayNeo)举办2026雷鸟AI眼镜新品发布会
在2026世界机器人大会主论坛上,明略科技创始人、CEO 兼 CTO 吴明辉带来题为《Agent+具身:应用需求牵引机器人智能的进化路径》的演讲,分享其在Agent、具身智能与产业…
机器人看3秒演示就能学会
对于大型企业而言,衡量AI Coding成熟落地的标志,不在于能写多少代码,而是能否在安全可控的前提下,实现组织级的规模化复制。 亚信深耕ICT与企业数字化领域30余年。近年来,亚…
雷峰网独家获悉,阿里云渠道生态业务近期完成人事调整,程传珂已接任渠道生态负责人,原负责人李鹏卸任。程传珂此前担任阿里云新生态华南大区总经理。 知情人士告诉雷峰网,程传珂在接手生态业…
汽车智能化竞争进入深水区,软件研发效率决定了产品迭代的速度,成为车企的核心竞争力。 相比通用软件研发,智能座舱研发更加复杂 —— 软硬件结合、多技术耦合、调用链路长,叠加多平台、多…
“4.8折?你确定?” 看到刚打探到的友商报价,云销售张小雨的第一反应是:不可能。 之前有智谱原厂销售跟他说,他们都是卖 8.5 折,现在友商竟然直接报出4.8折,比原厂最低价还低…
让科学家的时间回到科学创造
雷峰网独家获悉,快手可灵AI核心技术骨干王鑫涛离职。 王鑫涛本科毕业于浙江大学,博士毕业于香港中文大学多媒体实验室(MMLab),师从汤晓鸥等。他长期研究图像与视频生成,也是Rea…
2026年Q1,百度智能云增速达到79%,在中国头部云厂商中增速居首。在5月的百度开发者大会上,百度智能云副总裁、渠道生态部总经理尹英利在百度开发者大会上披露:有10%的核心伙伴业…
看完阿里昨晚发的最新财报,不想承认,却不得不承认的一个事实是:虽然市场聚光灯打在模型层和应用层,但短期来,AI 商业化最强的还是在云和芯片,而且是,当上层模型用得越凶,底层卖铲人越…
在全球脑科学快速发展的背景下,神经调控、人工智能、康复工程等方向正在不断取得突破。然而,围绕脑瘫,特别是由缺氧缺血性脑病(HIE)引发的长期功能障碍,如何连接基础脑科学、工程技术、…
近日,3D存算一体芯片领军企业谦合益邦宣布,其自主研发的全球首款4层3D DRAM堆叠存算一体芯片成功回片并点亮,标志着4层3D DRAM堆叠存算一体芯片从技术验证正式迈入工程落地…
8月19日,2026世界机器人大会在北京亦庄开幕。作为全球机器人产业的重要风向标,本届大会汇集了人形机器人、工业机器人及大量具身智能解决方案。在星炽动力展台,搭载PULSE具身智能…
离开谷歌的原因之一:小团队可以极致聚焦!
墨奇,开始加速了
<img class="rich_pages wxw-img" data-aistatus="1" data-croporisrc="https://mmbiz.qpic.cn/m…
<img class="rich_pages wxw-img" data-aistatus="1" data-croporisrc="https://mmbiz.qpic.cn/s…
<img class="rich_pages wxw-img" data-aistatus="1" data-croporisrc="https://mmbiz.qpic.cn/s…
优必选完成具身智能大模型备案,首批企业之一。
优必选展示人形机器人多领域应用成果。
当前,工业具身智能的规模化落地正卡在关键瓶颈:预训练模型在真实产线的部署成功率普遍不足60%,吞吐率不及人工,叠加居高不下的运维成本,让“机器人进厂干活”迟迟无法大规模兑现。 这一…
8月20日,网易有道(NYSE:DAO)发布了2026年第二季度未经审计的财务业绩。
Surging demand for AI training data is driving rapid growth for the startup and its rivals…
Businesses are willing to flop back and forth as each lab releases new models, volatility …
Ever wanted someone else to do your texting for you? ChatGPT is being offered up as an aut…
Jason Kelce joked that people should cool data centers with their pee, rather than potable…
Google is giving publishers a new button that lets readers make them a preferred source ac…
Linkdaze's smart digital calendar stands out for not putting its features behind a paywall…
Affected users told TechCrunch they were using Grok Lite, and noticed the issues as early …
ChatGPT and other AI models are now authoring and editing much of the new web.
Ramp has launched its own AI model routing service, dubbed Router, that lets users and com…
Meta is bringing Pocket, its experimental AI-powered app for creating and sharing interact…
Introducing AI Futures, a new OpenAI blog exploring how transformative AI could reshape po…
With a fixed deadline and design resources committed elsewhere, Stampli used Codex and Cha…

YouTube · NVIDIA