搜索“Agent”

rk . The question, according to a16z, is less about who wins and more about how well each company capitalizes on its window. The race for the personal assistant Six months ago, an agent like OpenClaw wasn't usable for most consumers, according to a16z. Now startup agents like Instinct and Tomo report hundreds of thousands of users, and the big players are countering with Meta's Muse, OpenAI's Dots, and xAI's Grok Bot. Agents become an economic force once they start triggering real purchases. Ins

匹配字段正文

The paper for MiMo-V2.6 by Xiaom is out. In MiMo-V2.6, AI runs much of its own training loop: agents build the tasks, audit the tests, grade the answers and hunt for cheats. Humans set the budget and the rules. shows that agent models kept improving with more RL compute by scaling batch size, task and harness variety, and grading effort together. Scaling RL for coding agents is hard because pass/fail tests can't tell a clean fix from a hacky fix. Agents also learn to game environments, for examp

匹配字段正文

s from a local government map’s settings file and a state agency’s dashboard. Claude Opus 5 and Mythos 5 also slipped past fetch-tool URL length limits, which guard against injection attacks, by using free link shorteners. Many cases began with ambiguous or impossible tasks, and Anthropic is fixing training environments that rewarded working around blockers. Public web benchmarks such as BrowseComp run on the live internet by default, so rival labs testing agents that way face the same exposure.

匹配字段正文

它的结构与“上帝手术刀”CRISPR高度神似，具备切割、复制、粘贴DNA的特征。 发现过程，竟动用了950个AI Agent，运行21小时，消耗约2.1亿Token。 随后，人类科学家接手审查和实验验证。 最终，团队进一步实验确认，这是一种此前未被表征的、存在于噬菌体中的全新酶系统。 他们将其命名为ART，即“阵列关联逆转录酶系统”。 CEO Dario难掩激动直言，“哪怕这是我在读博期间做出的成果，也足以骄傲一辈子”。 这一惊天发现的背后，是一场极其疯狂的算力风暴与算法协同。 在过去的生物学史上，寻找新的酶系统就像在干草垛里找一根特定的绣花针。 人类科学家需要耗费数月甚至数年，一行行比对碱基序列、查阅文献、排除噪音。 但在Anthropic的实验室里，这一流程被浓缩到了短短21小时—— 950个Claude集群并联启动，各司其职；它们吞吐、反刍、推演了超过2.1亿个Token的生物基因组数据。 然后，它们跨越数以万计的公开序列，地毯式扫描超20万个逆转录酶（RT）。 筛选出3500个全新候选系统，再层层严苛审视、推演排除，最终浓缩提炼出最核心的20个极具潜力对象。 在这个过程中，An

匹配字段正文

. We have updated the guardrails on some of our internet access tools, such as the web fetch tool, to heavily restrict what the model can do with them. And, as part of our ongoing agent security and monitoring efforts , we have built tooling to automatically detect and block the kinds of behaviors described above. This tooling now runs on most of our evaluations and on internal agentic use of frontier models. When we tested it against the cases described in this post, it blocked all of them. As

匹配字段正文

0x and inference time by up to 4x. We benchmark closed-source (OpenAI, Google) and open-source (Qwen) models on data extraction (CoNLL++ NER), multi-turn maze navigation (BabyAI), agentic RAG (Multi-Hop), and agentic tool use (τ-bench); Learn how to reproduce this workflow for your own use case using 11.5KTensorZero and other open-source LLMOps tools, even without a GPU, along with best practices for production deployments The Performance-Cost Dilemma Every LLM application builder faces a fundam

匹配字段正文

y used test, Needle in a Haystack (NIAH), is a simple lexical retrieval task often used to generalize a model’s ability to reliably handle long context. Real applications, such as agent tasks or summarization, demand significantly more processing and reasoning over broader, often more ambiguous information. Designing realistic long context benchmarks is challenging. Tasks often grow in complexity as input length increases, making it difficult to isolate whether performance drops are due to longe

匹配字段正文

quires a chain of intermediate searches in which the output of one search informs the next, a process known as a multi-hop retrieval. To solve this, leveraging LLMs for multi-turn agentic search has become a viable approach to answering multi-hop retrieval queries. Rather than issuing a single query, an LLM agent iteratively decomposes a high-level question into subqueries, retrieves evidence, and refines its search strategy across multiple turns. Concurrently, it has been shown that smaller-par

匹配字段正文

ks UV light, so it can only be measured from space. NASA's GALEX mission covered about two-thirds of the sky but skipped bright star-forming regions. Claude Science coordinates AI agents that downloaded data from multiple space missions, calibrated it, and merged it together. They filled missing areas using inpainting, a technique where a model learns from existing data to reconstruct gaps. In tests, predictions averaged about ten percent deviation from actual measurements. The map is meant to s

匹配字段正文

cant amounts of evidence in reasoning traces, logs, and other telemetry.1 This has made it much easier to notice when they misbehave. However, observability against an adversarial agent only helps if the agent can’t subvert that observability. When treating AI systems as potential adversaries, their transcripts, reasoning, actions, and other outputs should be considered untrusted input, and the systems that record and display them should be considered security-critical infrastructure. Given the

匹配字段正文

with students falling back on AI at the first sign of difficulty. That dynamic is now threatening one of MIT's most important programs. Faculty members are starting to consider AI agents instead of students as research assistants, which would hit the Undergraduate Research Opportunities Program (UROP), a pillar of the MIT experience: 93 percent of the class of 2025 participated at least once, and 58 percent of faculty served as mentors. The program exists to train students, not to provide cheap

匹配字段正文

vision-language models actually understand how the physical world changes over time? Modern vision-language models are increasingly being used in areas such as robotics, embodied agents, and video understanding. They can recognize objects, describe scenes, and answer surprisingly sophisticated questions. However, being able to describe a scene is different from maintaining an understanding of an object or physical quantity as the scene changes. We study this through the concept of conservation.

匹配字段正文

oss、reward等指标立即验证。 也正因为如此，Good当年那个听起来非常遥远的循环，今天才第一次开始有了比较具体的工程路径： 论文把这类路径称为software-driven intelligence explosion，也就是 由软件改进驱动的智能爆炸。 主要提一嘴的是，这和造出更快芯片还有点区别。 芯片设计、流片、建厂都有长周期，而算法、训练方法、Agent工作流、合成数据策略一旦验证有效，理论上可以很快重新部署进下一轮研发。 接下来，就是整篇论文最《抓马》的部分了。 假设一家实验室今天有1000名顶尖研究员，未来它开发出了一个能力达到顶级人类AI研究员水平的Agent。 从这一刻开始，扩张研发团队这件事，游戏规则就变了。 它还能24小时工作，成千上万个实例同时跑任务。底层模型一更新，整支AI研发团队也能同步升级。 这就直接绕开了人类科研最硬的一层增长限制： 人才培养速度 。 人类想把1000名顶尖研究员变成100万人，几乎是代际工程。 对此论文还专门引入了一个概念： effective R&D workforce，也就是 有效研发劳动力。 而且这几百万个研究员还有一个人类组

匹配字段正文

献，与去年基本持平。; 39%的受访者预计未来一年AI将导致所在企业裁员，高于去年预期，但预期兑现率有待观察。 麦肯锡发布的2026年度报告《麦肯锡全球AI应用现状调研》（McKinsey Global Survey on the State of AI）覆盖97个市场，汇集了超过1700位高管的反馈。结果显示，AI的应用范围正在持续扩大，尤其是AI智能体（Agentic AI）的规模化部署较一年前明显提速，而少数 “高绩效企业”已初步展现出规模化部署AI的实际成效。 对于中国科技、制造与金融行业的商业领袖而言，这些调研结果更像一份行动清单，而非警报。中国有庞大的制造业规模、成熟的数字基础设施，以及快速适应全球技术周期的深厚积累。这些优势使中国企业有能力将调研所揭示的方向付诸实践，前提是企业领导者具备与高绩效企业同等的战略纪律。不过本次调研未单独披露中国市场的分项数据，此处关于中国企业的判断基于全球趋势与中国产业结构特征的推断。 01 规模化部署：从零散试点到全面推行 这一阶段已接近尾声：此次调研中，近九成受访者表示所在企业至少在一项业务职能中常态化使用AI；已经跨越单点部署、迈向全企

匹配字段正文

A tool call is not an outcome Final responses and valid tool calls are only proxies. An agent can sound correct while leaving the wrong value, changing the wrong record, or creating an extra side effect. Only the records it leaves behind settle the question. The gap is substantial. In a common-set ablation covering 121,680 valid trials across 12 LLM models, 79,853 attempts failed the executable checks. Of those failures, 67.24% still terminated cleanly, invoked a state-changing tool, and reporte

匹配字段正文

Coding agents can build an editable 3D scene from a single photo by writing and refining code step by step. The benchmark that comes with this work reveals where the process still breaks down: self-assessment and geometric accuracy. A 3D reconstruction from a single photo is most useful when it exists as an executable program you can inspect, edit, and query. That's the premise behind LEGO-Anything, a project from researchers at the University of Maryland and AWS. The approach is called "Image-t

匹配字段正文

zon benchmark built to measure learning from real-world environments . The benchmark comprises 134 realistic and diverse tasks spanning six major capability domains, each allowing agents to operate continuously for at least 12 hours. Based on long-horizon interaction runs across these tasks, we find that the overall agent performance in "environment learning" closely follows a log-sigmoid curve , with an average goodness of fit of R² = 0.998. Across frontier models from different generations, ag

匹配字段正文

rts Weigh the Risks, Rules and Race Ahead Shana Lynch Sep 22, 2026 News In a new series called Prompt Response, Stanford’s Surya Ganguli, Diyi Yang and Rob Reich examined emergent agent behavior, recursive self-improvement, independent evaluation and whether a kill switch can make advanced AI safer. In a new series called Prompt Response, Stanford’s Surya Ganguli, Diyi Yang and Rob Reich examined emergent agent behavior, recursive self-improvement, independent evaluation and whether a kill switc

匹配字段正文

ew Stanford research finds they often don't measure what they claim to. In a new series called Prompt Response, Stanford’s Surya Ganguli, Diyi Yang and Rob Reich examined emergent agent behavior, recursive self-improvement, independent evaluation and whether a kill switch can make advanced AI safer. Good Morning America Sep 21, 2026 Media Mention HAI Policy Fellow Jennifer King advises caution surrounding one's person security when using AI, citing her research on AI privacy and insights on user

匹配字段正文

rts Weigh the Risks, Rules and Race Ahead Shana Lynch Sep 22, 2026 News In a new series called Prompt Response, Stanford’s Surya Ganguli, Diyi Yang and Rob Reich examined emergent agent behavior, recursive self-improvement, independent evaluation and whether a kill switch can make advanced AI safer. In a new series called Prompt Response, Stanford’s Surya Ganguli, Diyi Yang and Rob Reich examined emergent agent behavior, recursive self-improvement, independent evaluation and whether a kill switc

匹配字段正文