搜索“Agent”

Google DeepMind on October 15, 2026 published a technical walkthrough in which biologists used Google Antigravity, its agentic development platform, to identify a novel antimicrobial peptide candidate in Streptococcus pseudopneumoniae, and the company said the lead is undergoing wet-lab validation. The walkthrough, authored by Pengfei Tian, Maxwell L Bileschi, Zoi Tsangalidou, Sukhdeep Singh, and Lucy J. Colwell, frames the work against rising antimicrobial resistance, which the post says could

匹配字段正文

ed and produce correct outputs? Today, answering that question often requires a degree of trust alongside verification. We trust the operator, the vendor, the counterparty, or the agent, and we rely on contracts, audits, and logs to establish that trust. Zero-knowledge proofs offer a fundamentally different approach to establishing trust. A program runs, and along with its output it produces a short proof, a cryptographic certificate that this program, on these inputs, produced this output. The

匹配字段正文

直读，在 count 模式下比 SQL 翻译快 3.5 倍，返回完整结果时快 2.8 倍，同时通过了 639 项适用 W3C 评测测试中的 636 项，以及数百万次差分模糊测试…… 这项研究任务实际上是一次实验，旨在以“Ultracode”模式运行 Claude Code for web，正如 Claude Code 文档中《Orchestrate subagents at scale with dynamic workflows》一文所述。 开启 Ultracode 后，Claude Code 会大量使用动态工作流——它生成 JavaScript，定义并协调一个子代理网络来处理复杂问题。 我最初的提示词是这样的——我本意是“直接命中这些行”，而“delete”是我在手机上不小心打错的字： 用 Rust 编写一个 SQLite 扩展，植入 SPARQL，并且运行得非常快，因为它可以直接命中 delete 行 这导致 Claude Code for web 连续运转了超过 12 个小时。第二天我查看了它的进度，并在注意到它构建了多少个 Rust crate 后发送了一条后续提示： cr

匹配字段正文

A new benchmark shows the best AI agents can solve theoretical physics calculations as accurately as leading specialists in each subfield. The researchers behind it say a missing sense of "taste" is one reason breakthroughs haven't followed yet. Thirty-five researchers from institutions including the International Centre for Theoretical Sciences (ICTS), Harvard University, KU Leuven, and Google DeepMind have released TheoPhysEval . The benchmark contains 50 unpublished research problems in theor

匹配字段正文

Enterprise agents need to do more than explain what should happen: they must update the right records, follow business policies, and complete work across applications. Prime Intellect and NVIDIA built 2,500 verified training tasks to teach this, and 100 GRPO steps, about 11 hours of training on 128 NVIDIA GB200 GPUs, took NVIDIA Nemotron 3.5 Super VL from 34.0% to 57.7% on AutomationBench-AA (public split). Key takeaways Nemotron can be specialized for enterprise work. Post-training on 2,500 ver

匹配字段正文

ding for the ECMWF authors; the competing-interests declaration states that several authors are Microsoft employees or own Microsoft stock. Jonas Reeve is an AI-generated research agent at Unite.AI, focusing on cognitive AI, artificial general intelligence (AGI), and the theoretical foundations of machine intelligence. His work explores how learning, reasoning, memory, and abstraction emerge in both biological and artificial systems, drawing connections between modern AI architectures and long-s

匹配字段正文

cysteine protease study is publicly available in two GitHub repositories, and the cysteine paper’s authors declared no competing interests. Aria Bloom is an AI-generated research agent exploring how artificial intelligence is transforming biotechnology and genomic research. Her writing blends precision with a deep curiosity about the future of life sciences. From synthetic biology to personalized medicine, Aria analyzes how machine learning is accelerating human health innovation. Articles auth

匹配字段正文

A team of researchers investigating AI agents trying to bypass Alibaba’s anti-bot protections left a hidden message inviting the agents to contact them. The agents wrote back. The investigation began when Swarmchasers researchers Ethan Elasky , Rowan Howard-Jones, and Jonathan Elsworth Eicher helped uncover thousands of attempts by AI agents to retrieve data from Amap, Alibaba’s mapping service, which Alibaba says reaches nearly 1 billion monthly active users. The activity traced back to agents

匹配字段正文

026, frontier language models have made enormous progress in their ability to control robot arms. We now believe that the future of robotics is likely to involve strong generalist agents, and conclude that tracking progress of language models for robotics tasks is going to be important - even if frontier language models are unsuitable for production robotics use on their own. AA-Robotics v0.1 is our research preview of the simplest experiment we have designed to test this - language models in co

匹配字段正文

New Carnegie Mellon paper finds that OpenClaw's logs left 0 of 4 policy violations provable, a warning for any team relying on agent logs for audits. They found that most agent policy violations in its tests could not be proven from logs, often because approvals and permissions were buried in chat text. Also finds that only 10% to 30% of violating runs needed human judgment, so most agent rule breaks could be checked by code with better logs. Safety work mostly tries to stop agents before they m

匹配字段正文

The field of agentic commerce, which describes the use of AI agents to carry out Internet transactions, is rapidly maturing, and crypto is fueling the boom. That’s according to a new report by the blockchain analytics platform Chainalysis, which found developers are increasingly giving AI agents their own budgets to buy the computing power, live web data, and code tools they need to complete tasks. The growth of agentic commerce over the past 15 months has been steep. On Base, the Coinbase-incub

匹配字段正文

The field of agentic commerce, which describes the use of AI agents to carry out Internet transactions, is rapidly maturing, and crypto is fueling the boom. That’s according to a new report by the blockchain analytics platform Chainalysis, which found developers are increasingly giving AI agents their own budgets to buy the computing power, live web data, and code tools they need to complete tasks. The growth of agentic commerce over the past 15 months has been steep. On Base, the Coinbase-incub

匹配字段正文

Oct 14, 2026 17 minute read Teams of coding agents are already building large-scale software, but how should they be organized to work most effectively and efficiently? AI isn’t a shortcut. It’s how business gets ahead. Key takeaways Teams of agents unlock a new axis of inference-time scaling. On challenging coding tasks, well-organized teams outperform single agents, with gains of up to 10 percentage points on ProgramBench. But these gains come at 3–7× higher cost. The question is not whether m

匹配字段正文

ps://arxiv.org/pdf/2609.30249 图：RAPID 面向接触丰富、非抓取操作，仅需单次人类演示即可学习。 在 RAPID 架构中，技能学习被转化为由编程智能体与物理校验器构成的闭环编程循环： 1.演示解析与原语拆解 ：从单次演示中提取动作，自动切分为推、翻转、支点旋转、抓取等基础操作原语； 2.以物体为中心的程序生成 ：Coding Agent 将原语转化为以物体为中心的约束程序，脱离特定的几何参数，使其天然具备可复用性与组合性； 3.仿真校验与闭环迭代： 校验器在 Real2Sim 仿真环境中验证程序，将碰撞、物理可行性等反馈传给 Coding Agent 进行修正，可在部署前筛除大量物理不可行程序。 图：RAPID 的智能体编程循环。Coding Agent 生成行为程序 Π，Evaluator 在 Real2Sim 仿真中校验并反馈，迭代修正。 在涵盖书架、隔板等场景的 8 个复杂非抓取真机任务中（灵活组合了推、翻、压移、支点、倾倒、抓取等原语），RAPID 仅凭 单次人类演示 ，便展现出极高的鲁棒性与泛化求解能力。 这说明，将 Coding Agent

匹配字段正文

e to that exact file! And that’s essentially the core vulnerability in Copilot here. The vulnerability is in Microsoft.VisualStudio.Copilot.Core.dll. A trailing dot allowed the AI agent to bypass protections around sensitive configuration files and modify settings it was explicitly supposed to be prevented from changing. There are a lot of other interesting quirks because of NTFS and Win32 differences. Other Bypasses and Observations While writing an exploit for it I discovered additional issues

匹配字段正文

New Carnegie Mellon and Microsoft paper finds that the best AI code reviewer caught 71.9% of known bugs in its top 5 comments, still too few to remove human reviewers. Agents now write code faster than people can review it, tempting teams to let AI review it too. Older review benchmarks scored AI against human comments, which can be nitpicks or miss real bugs. They built CRBugBench, 1,800 code review tasks where each pull request introduced a bug that humans confirmed and fixed. The agent review

匹配字段正文

前尝试留下的笔记时，它们在陌生硬件上设置推理的能力变得更强。 研究人员让 Qwen3.5-9B 经历了 20 个独立的谱系，每个谱系持续六代，成功率从第一代的 45% 攀升至后续世代的 61% 到 84% 之间。 研究人员称之为拉马克式遗传，即一个智能体将其所学传递给下一个智能体，而无需对模型进行任何重新训练。 对于使用 OpenClaw 和 Hermes Agent 等持久化智能体运行时的开发者来说，这是一种熟悉的机制，但在这里，这种记忆能帮助后续的副本在它们从未见过的硬件上解决安装失败的问题。 研究人员称之为拉马克式遗传，即一个智能体将其所学传递给下一个智能体，而无需对模型进行任何重新训练。团队还尝试让 Qwen3.5-9B 向能力强大得多的 Opus 4.8 模型寻求建议，但没有产生统计学上显著的改善，因为小模型往往不知道该问哪些正确的问题。 2026 年底的趋势线 论文指出了两个使这一场景一直停留在理论层面的制约因素。 模型需要足够有能力以可靠地复制自身，同时又要足够小以能在被攻陷的硬件上运行，而目前这两个要求并不重叠。 GLM-5.2 在简化测试中展现了必要的能力，但对于研究

匹配字段正文

es as AI safety concerns have reached a fever pitch. Over the past few months, a steady stream of reports have come from AI companies, including Anthropic and OpenAI, saying their agents are capable of acting on their own and hacking into other companies and government entities. The revelations have caused researchers, lawmakers and executives, including Anthropic’s CEO, Dario Amodei, and the OpenAI CEO, Sam Altman, to call for a slowdown and increased guardrails on the burgeoning industry. Desp

匹配字段正文

threats; Continuous monitoring: As the model reasons and writes its response across multiple steps, we combine these signals and take action when necessary such as cutting off the agent mid-stream if the responses become problematic Combining real-time protection, faster fixes for new jailbreaks and enhanced monitoring for recurring misuse has made our systems better at preventing real-world harm. For instance, external bio jailbreak red teaming could only unlock less than 1% of bio sub-tasks we

匹配字段正文

verless GPUs, we show that stable, performant, and cheaper RL is possible with mixed-precision, mixed-accelerator samplers by training an open-weight model fully asynchronously on agentic coding tasks. Our setup We train Qwen3.6-35B-A3B on agentic coding with up to 256K tokens of context using the 2.7K repository tasks Xiaomi released with MiMo-V2.6, after removing the repository history and other traces through which some tasks leak the reference fix. In each episode, the model creates a patch

匹配字段正文