Skip to main content

XMT

短闻

信流 · 上滑连读 · 来源可核

今日 稍后 搜索 RSS
1 / 24
Aggregate arXiv cs.AI 人工智能 45″

Fool's Gold: Defensive Deception Against Safety-Removal Attacks on Open-Weight Models

arXiv:2608.…

  • 17202v1 Announce Type: new Abstract: Safety alignment in open-weight l…
  • What cannot be prevented can be deceived.
  • Our defense, decoy hardening ("Fool's Gold"), concedes the refusal str…

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate arXiv cs.AI 人工智能 45″

Benchmarking the Benchmarks: Evaluating Automated Safety Benchmarks for Small Language Models

arXiv:2608.…

  • 17183v1 Announce Type: new Abstract: Small Language Models (SLMs) are …
  • However, existing AI safety\slash security\slash compliance benchmarks…
  • We therefore ask: Can these benchmarks effectively and reliably evalua…

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate arXiv cs.AI 人工智能 45″

Synthesizing Feature Extractors: An Agentic Approach for Algorithm Selection

arXiv:2608.17170v1 Announce Type: new Abstract: Algorithm selection for constraint satisfaction problems requires extracting features that capture problem structure.…

  • Manually designing feature extractors demands deep domain expertise an…
  • We present an automated approach that uses Large Language Models (LLMs…
  • Given a high-level MiniZinc model and an instance, the LLM agent gener…

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate arXiv cs.AI 人工智能 45″

KnowSim: Evaluating Information Calibration in LLM Assistants with User Simulators that Learn

arXiv:2608.…

  • 17150v1 Announce Type: new Abstract: To effectively collaborate with u…
  • Yet user simulators used to evaluate and train LLMs do not explicitly …
  • To close this gap, we introduce KNOWSIM, an evaluation framework built…

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate arXiv cs.AI 人工智能 45″

Toward Personal Intelligence Through Cooperative Observation

arXiv:2608.…

  • 17128v1 Announce Type: new Abstract: A personal AI system needs a mode…
  • Broader observation does not by itself improve assistance because a bo…
  • We argue that this observation bottleneck has a cooperative structure:…

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate WWD 时尚 13″

Trump Pauses 50% Tariffs on Canada For Three Days With Deal In Sight

About an hour before the deadline Wednesday morning, President Donald Trump Truthed that he had "paused" the trade action for a three-day period.

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate InfoQ 中文 人工智能 6″

Valhalla 项目首次预览:JEP 401 重新定义 Java 对象的 == 运算符

点击查看原文>

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate 极客公园 人工智能 45″

马斯克带火的太空算力领域,传一个 800MB 模型上天要四周

「所有劝我不要干太空算力的朋友,现在全在干太空算力。」 太空字节联合创始人秦贵军这样描述过去一年。这家公司 2024 年底成立,做的是太空数据中心;他本人参与过北斗导航重大专项,在卫星行业干了十几年。…

  • 那些劝过他的朋友也大多是同一个画像:传统航天出身,两年前还觉得把算力送上天是个不着调的故事,如今已经用脚投票。
  • 他画过一张国内太空算力的行业热力图,从长三角到北京亦庄、中关村,再到成都、西安、合肥,玩家已经超过三十家,正在往四十家走。
  • 2026 年 6 月 12 日,SpaceX 以每股 135 美元、约 1.

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate arXiv cs.AI 人工智能 45″

A decodability criterion predicts when hidden-state selection beats majority voting in large language models

arXiv:2608.…

  • 17124v1 Announce Type: new Abstract: Combining the answers a large lan…
  • Voting is unreliable on difficult questions, where the sampled answers…
  • Selecting a candidate by reading a correctness signal from the model's…

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate arXiv cs.AI 人工智能 45″

DiSCO: Defending text-to-image generation through distribution-guided contrastive prompt optimization

arXiv:2608.…

  • 17067v1 Announce Type: new Abstract: As text-to-image generative model…
  • Existing defenses predominantly operate under white-box assumptions, r…
  • Black-box alternatives based on LLM prompt rewriting offer broader app…

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate arXiv cs.AI 人工智能 45″

KernelArc: A Multi-Agent Framework for GPU Kernel Optimization

arXiv:2608.17071v1 Announce Type: new Abstract: We present KernelArc, a multi-agent framework for autonomous GPU kernel optimization across heterogeneous workloads.…

  • Strategy-specialized agents run in parallel and coordinate through con…
  • We evaluate \kernelarc{} on NVIDIA H100 and B200 GPUs using category-r…
  • The resulting implementations span custom BF16 GEMM, static cuBLASLt E…

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate arXiv cs.AI 人工智能 45″

Memory Is Communication: The Frontier Between Remembering and Signaling

arXiv:2608.17053v1 Announce Type: new Abstract: A bounded agent may obtain information for a decision from its own past, from peers, or from both sources.…

  • Retaining task-relevant history can reduce later communication, while …
  • Under limits on both resources, how should an agent allocate its infor…
  • Given a fixed task and decision rule, the memory and message rate pair…

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate arXiv cs.AI 人工智能 45″

SkillEffect: Checked Lowering for Memory-Bounded Agent Tools

arXiv:2608.…

  • 17007v1 Announce Type: new Abstract: Agent Skills can specify procedur…
  • However, when models turn this guidance into code for existing tool in…
  • We present SkillEffect, a checked-lowering runtime for computations wi…

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate 数英网 时尚 45″

文章频道 - 上海书展手写推荐语,一句顶一万句

写出推荐语的编辑太懂营销了。只要AI还在,手搓就会被偏爱。《牛来》之外,这届上海书展凭借手写推荐语,小小出圈了一次,再次见证文案的力量。编辑把读完一本书后最想说的话写在便利贴上,再贴回书封。…

  • 方寸之间的几行字,让不少读者停下来翻书,一些原本有阅读门槛的历史著作也在现场售罄。
  • 比如这句“探险家斯坦因盗猎中国文物的罪证!
  • ”,让冷门书都卖完了没有介绍作者生平,也没有概括章节内容,几个带有情绪和冲突的词,已经给出了打开这本书的理由。

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate Hypebeast 时尚 45″

Carnan and ASICS Reunite for a Purple-Draped GEL-CUMULUS 16

Name: Carnan x ASICS GEL-CUMULUS 16 “Path II”Colorway: “Path II”SKU: TBCMSRP: TBCRelease Date: August 19Where to Buy: Carnan, ASICS BrazilFollowing up on the success of their inaugural partnership back in 2024, Brazilian label Carnan and ASICS have reunited to unveil the second chapter of their collaborative "Path" project.…

  • The Carnan x ASICS GEL-CUMULUS 16 “Path II” sneaker retains the black …
  • A deep, muted purple hue now dominates the upper, representing the flo…
  • A brown insole connects back to the roots of the collaboration, reinfo…

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate WWD 时尚 8″

Country Artist Ian Munsick and Schaefer Outfitter Create Westernwear Capsule

The launch of the collection coincides with the drop of his new album, "The Mountain Goat."

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate WWD 时尚 10″

A&F Expands Distribution of NFL Product Through Fanatics Partnership

Its men's, women's and kids' collections for all 32 teams will be offered in all of the league's commerce channels.

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate WWD 时尚 18″

Snoopy in Style to Open Exhibition in New York’s Meatpacking District on Sept. 2

A spectrum of brands — including Coach, Michael Kors Collection, Christopher John Rogers, Tommy Hilfiger, and GapStudio designed by Zac Posen — have designed custom-made pieces for Snoopy and Woodstock plush dolls.

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate WWD 时尚 13″

Inside the WWD Archives: When New York’s Art World Entered Fashion’s Orbit

A look at how Pop Art's influence on fashion and artists Andy Warhol and Roy Lichtenstein highlighted the enduring relationship between the two realms.

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate 少数派 综合科技 6″

App+1 | Inkive:把纸质书上划好的线存进 Obsidian

让纸书上的划线,回到你的数字笔记里。查看全文

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate CleanTechnica 新能源 32″

Solar Power Passes 3 Terawatts!

Solar power has passed another big milestone.Solar has reached and quickly surpassed the 3-terawatt — yes, terawatt — marker globally.…

  • That’s how much solar power is installed across the world.
  • Naturally, a lot of the growth was from China, but a lot of it was als…
  • [continued] The post Solar Power Passes 3 Terawatts!

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate 爱范儿 人工智能 45″

OpenAI 紧急暂停新模型训练,AI 开始进入「越聪明越危险」阶段

今天凌晨,OpenAI 突然宣布,在 Hugging Face 入侵事件后已经暂停了最新模型的强化学习(RL)训练两周;同时,原计划进行的迄今最大规模前沿强化学习训练也处于暂停状态—— 他们计划用这段额外的时间评估模型行为、获取有关模型安全对齐状况的信息并调整安全措施。…

  • 当然,这也意味着预热已久的下一代大模型 Astra 大概率又要跳票了。
  • 在大模型「卷生卷死」的时代,为什么 OpenAI 要主动踩刹车?
  • 归根结底,还是安全问题。

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate 雷锋网 人工智能 45″

支付宝“阿宝”跨端服务落地16家车企 新增60余家定点合作

8月17日,国内首届智能体商业生态合作伙伴大会于杭州正式举办。支付宝携手比亚迪、吉利汽车、理想、蔚来、东风汽车、深蓝汽车、上汽乘用车、一汽奔腾等主流整车企业,联动千问、华为、OPPO等大模型及终端头部厂商,共计20余家生态伙伴,共同发起多智能体互联协同生态共建计划,加速智能体商业从技术概念走向规模化商用落地。…

  • 本次生态共建,各方将依托各自技术积淀与产品优势,探索多智能体可信协同标准化路径,彻底打破跨设备、跨主体的智能服务孤岛。
  • 通过统一的智能体联动机制,用户可依托自然语言交互,在不同智能终端、不同厂商智能体之间,获得连贯、安全、闭环的办事与交易服务,全面升级全场景智…
  • 大会上,支付宝正式发布国内首个AHA多智能体跨端互联协议体系。

RSS 官方收录 · 可信分层展示

详情 原文 分享图
Aggregate arXiv cs.AI 人工智能 45″

The Problem Is the Problem: Towards Scalable Mathematical Discovery

arXiv:2608.16977v1 Announce Type: new Abstract: AI systems are increasingly capable of contributing to mathematical research.…

  • In research practice, frontier-model reasoning is a limited resource, …
  • Allocating these scarce resources well is therefore central to making …
  • In most current AI-for-math workflows, human effort is concentrated at…

RSS 官方收录 · 可信分层展示

详情 原文 分享图