Skill Self-Play: Pushing the Frontier of LLM Capability with Co-Evolving Skills· 技能自我对弈:用共进化技能推动LLM能力边界
LLM training is shifting from manual design and annotation to interaction-driven self-evol…
LLM training is shifting from manual design and annotation to interaction-driven self-evol…
To effectively integrate AI into high-stakes, critical environments such as healthcare, au…
Adding procedural skills to an LLM agent is typically evaluated by average improvement in …
Routing to select large language models (LLMs) with different cost-quality trade-offs has …
Existing lexical distance, similarity, and evaluation metrics operate on Unicode code poin…
Enterprise AI agents are typically granted static credential sets at configuration time, h…
Vision-language model (VLM) agents increasingly use tools to act on 3D scenes rather than …
Diagnosing the root cause of anomalies is essential for safe industrial operation. Despite…
Most endangered languages lack the parallel data required for machine translation, despite…
Large Language Models (LLMs) have significantly automated the process of scientific discov…
Agent benchmarks increasingly evaluate repository editing, web research, terminal use, and…
High-complexity operational environments require methods that detect and anticipate tempor…
Pluralistic value alignment---the goal of building AI systems that represent and serve div…
We present a systematic study of healthcare-domain cross-lingual transfer to address the s…
Reuse-based development has become increasingly important in the creation of complex syste…
Despite the growing use of large language models (LLMs) as creativity evaluators, evidence…
Asynchronous reinforcement learning (RL) accelerates large language model (LLM) post-train…
AI agents encounter learning opportunities in every episode they run, and discard nearly a…
Industrial health management increasingly relies on heterogeneous information sources, inc…
Large reasoning models (LRMs) generate long reasoning traces before producing final answer…
We present Nanbeige4.2-3B, a compact general agentic model with 3B non-embedding parameter…
Multimodal Large Language Models (MLLMs) are emerging as core reasoning modules for embodi…
Reinforcement learning with verifiable rewards (RLVR) has emerged as a highly effective fr…
Large language models (LLMs) have become powerful tools for language understanding and log…
TRACE-RealWorld addresses a core data-management problem: maintaining an actionable materi…
Approving is an open-source, self-hostable platform for turning coding agents into visual,…
A map of what AI tokens actually cost, and where they're wasted vs. well spent. Tools, res…
Since launching last year, the startup today has more than 8 million people using the open…
Recursive’s $400 million outlay represents the bulk of the company’s fundraising to date.
These AI features can help you craft a menu, design a tablescape, and handle other party-p…
It might sound counterintuitive, but Search's AI tools can actually help you make the most…
36氪获悉,立昂技术公告,公司发布2026年度以简易程序向特定对象发行股票预案,拟募集资金总额不超过2.55亿元,扣除发行费用后用于立昂云数据人工智能算力资源池项目及补充流动资金。
7月28日,亚马逊云服务(AWS)宣布与人工智能公司Recursive达成一项价值4.1亿美元的多年期协议。Recursive将指定AWS作为其云服务提供商,并在亚马逊云服务平台上…
英伟达签署最高价值500亿美元的巨型得州数据中心租赁协议。这项此前未对外披露的合作,再度引发市场关注:这家芯片巨头正不断深入参与AI基础设施融资环节。五位知情交易人士透露,这家市值…
7月27日,东莞市数字人体与智慧睡眠创新联合体揭牌暨工作推进会在慕思健康睡眠股份有限公司(以下简称“慕思”)总部举行。本次工作推进会由东莞市科学技术局、厚街镇人民政府指导,慕思、广…
作者 | 黄楠 编辑 | 袁斯来 硬氪获悉,人形机器人基础模型公司德塔智能(Delta Intelligence)近日完成近5亿元天使++轮融资,本轮投资方包括多家上市公司产业方和…
微软于周一发布全新人工智能安全产品 “感知计划(Project Perception)”,依托多款专为网络安全场景打造的自研 AI 模型驱动。微软将这款产品定位成 Anthropi…
把搜索能力开放给Agent了
<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…
7月28日,360集团创始人周鸿祎正式发布新一代企业智能体工作平台“纳米Work”。纳米Work为企业AI化而生,面向企业老板、创业者及一人公司(OPC)和企业员工,打造超级组织。…
7月28日,百度萝卜快跑联合隶属 Lyft 集团的Freenow,在伦敦启动公开道路测试,计划2027年起面向市民提供无人驾驶出行服务。这是中国无人车首次在英国伦敦复杂路况下实地路…
文|邓咏仪 编辑|张雨忻 吴秉哲每天至少复盘一次。 作为北大计算机博士,他同时也是一个高频的投资者。每天盘后,他会回顾当天的判断——哪些预案执行了,哪些被盘面的新信息打乱了,哪些潜…
20 26 年 7 月 27 日, 深圳 —— OPPO 宣布启动「小布 Next 计划」,开放行业首个端侧Multi-Agent协同系统内测。「小布 Next」 搭载行业首个端侧…
中国开源模型在海外取得显著成绩。
据知情人士透露,OpenAI首席执行官萨姆・奥特曼预计将于当地时间周三、周四重返华盛顿,与白宫及美国政府其他官员、国会两党议员举行会谈,讨论人工智能相关政策。其中一名消息人士透露,…
从“工具效率”到“认知理解”,教育AI进入深水区
<section data-mpa-powered-by="yiban.io" data-pm-slice="0 0 []" style="margin: 0px 0px 24px…
7月28日,阿里Qoder上线国内首个实时语音交互智能体Qoder Voice,用户语音唤醒智能体后,智能体可像真人助手一样实时与用户互动并执行任务,让用户解放双手,实现“动动嘴,…
2026年 , 教育部等五部门印发《“人工智能+教育”行动计划》,文件明确要求分层提升教师AI数字素养,全面构建人机协同的新型课堂模式。政策方向已然明朗,但落地层面的讨论仍在发酵。…
<img data-src="https://static.leiphone.com/uploads/new/images/20260728/6a684ad5e1b23.jpg" …
迷人的Scaling曲线
视频训练模型新增触觉感知能力。
日前,俗称“抱抱脸”的人工智能开源社区,遭遇OpenAI智能体入侵。当地时间26日,“抱抱脸”联合创始人兼首席执行官向OpenAI 提出两项要求,一是公开智能体全部轨迹,二是将1亿…
记者今天(7月28日)从雄安新区了解到,“人工智能大模型IPv6能力提升专项行动”由中央网信办联合北京、上海、浙江、深圳四地网信办,会同5家头部大模型企业在雄安新区共同启动,推动生…
扩散模型首次打通长程Agent任务
IJCAI-ECAI 2026将于8月15日至21日在德国不莱梅举办,990篇录用论文创近年新高。本期提炼出三大最热议题:Agentic AI领域两大最高荣誉获奖演讲同题共振,Ph…
36氪获悉,近日,灵光App上线闪应用“一键部署”功能,并向所有用户开放。用户使用Codex、Claude Code、Cursor、Qoder、Trae、Workbuddy等AI工…
36氪获悉,《杭州市人工智能产业发展促进条例(草案)》公开征求意见。其中提出,市人民政府应当统筹规划人工智能基础设施体系建设,适当超前布局算力网、新型电网、新一代通信网、数据可信空…
36氪获悉,《杭州市人工智能产业发展促进条例(草案)》公开征求意见。其中提出,拟支持词元经济布局发展。支持企业建设词元生产平台、交易结算平台、对外服务平台,提升本市的词元生产、调度…
针对人形机器人目前进展情况,比亚迪官方回应称,计划八月在“迪空间”亮相发布。此前比亚迪曾在“迪空间”发布一张人形机器人宣传海报,海报称“八月初,有个新朋友,想来认识你”。 (财联社…
Anthropic首席执行官达里奥·阿莫代伊当地时间7月27日发表文章称,Anthropic从未主张全面禁止开放权重AI模型。他表示,具备一定安全性的开放权重模型能够扩大AI技术获…
36氪获悉,7月27日晚,月之暗面发布Kimi K3的模型权重、技术报告,并开源支撑Kimi K3模型训练的关键Infra技术:MoonEP、FlashKDA和AgentEnv。据…
个人化AI应用趋势分析
7月28日,阿里云真武M890超节点实例已Day0适配Kimi K3,这是国内首个跑通近3万亿参数模型的超节点。双方将进一步展开国产算力合作,千问AI平台和阿里云百炼也将提供Kim…
20个月「像过完了一辈子」
36氪获悉,据阿里云消息,阿里云真武M890超节点实例已Day0适配Kimi K3,这是国内首个跑通近3万亿参数模型的超节点。双方将进一步展开国产算力合作,千问AI平台和阿里云百炼…
36氪获悉,国金证券研报指出,mSAP工艺适用于线宽线距要求高的应用场景(线宽/线距可达15um以下),技术难点环节主要在基铜、电镀、闪蚀等。mSAP最初应用场景主要为消费电子,A…
Anthropic founder and CEO Dario Amodei made his views clear about open-weight models and C…
在长鑫科技上市首日大涨的背后,参与网下配售的公募基金成为重要受益者。按上市首日收盘价测算,公募基金打新长鑫科技的总浮盈超500亿元,获配金额最高的基金公司浮盈超过68亿元;从单只产…
在多重考验交织下,银行业走完2026年上半程。随着年中工作会议的陆续召开,多家上市银行下半年的经营路线图也正式浮出水面。透过一系列会议可见:多家银行明确将加强资产负债管理、提升多元…
信用卡获客的开卡礼,正从传统的拉杆箱、咖啡券等,向人工智能(AI)算力、大模型会员等加速转变。2026年下半年以来,多家商业银行密集上线搭载AI权益的主题信用卡。“消费即算力”,正…
芯片巨头英伟达正与OpenAI进行一项规模空前的融资谈判,计划提供约2500亿美元的融资担保,以支持后者租用位于美国俄亥俄州南部的一个巨型数据中心园区。该园区由软银集团旗下能源子公…
Companies without their own models — or without a layer of AI infrastructure known as AI g…
The issue appears to have originated from Claude’s “share chat” feature, which allows user…
Microsoft bolstered its AI cybersecurity offerings this week with the launch of its first …
OpenAI's Hugging Face breach has reignited debate over AI alignment and control, exposing …
Meta on Monday said it is rolling out its Meta AI chatbot within Threads' DMs, giving user…
Google’s AI Overviews now appear in 43% of searches, underscoring how quickly AI-generated…
At TechCrunch Disrupt 2026, the Smart Systems Stage will be where energy, infrastructure, …

YouTube · OpenAI

YouTube · OpenAI

YouTube · OpenAI

YouTube · OpenAI

YouTube · NVIDIA