搜索“Agent”

discovery. This is especially true for neglected diseases such as animal envenomings, which have been largely starved of innovation due to low commercial incentives. Being able to agentically tackle the uniquely broad target space across animal toxins provides a significant speed-up in discovery, and brings realistic therapeutic solutions much closer. Timothy P. Jenkins , Associate Professor, Technical University of Denmark. Claude ran the design campaign with considerable autonomy Our initial p

匹配字段正文

Meta 正在把具备较强自主执行能力的 AI Agent，从云端大模型压缩到可以在个人电脑上持续运行的 Muse Glimmer 30B。 部分指标超过 Qwen3.6-27B 和 Gemma4-31B。 通过 4-bit 量化，将模型压缩到约 17GB，目标是在 24GB 显存设备上运行。 采用 DFlash 推测解码，在 RTX 5090 上实现约 233 tokens/s，相比普通解码提速约 3.1 倍。 Meta 最近牛逼克拉斯！

匹配字段正文

New Princeton paper finds that combining specialized agents from many developers can beat any single developer's general agent, with GAIA accuracy reaching 70%. Today most agentic systems are built and controlled by 1 developer, so each team rebuilds the same basic skills. The authors argue this limits scale. They simulated populations of developers with LLM agents that build agents under a fixed budget, then compared lone systems with a multi-developer system. The multi-developer system splits

匹配字段正文

tematic effort to map the full landscape of clinical tasks and measure how today’s best models perform across them. Then we went deeper with PhysicianBench, which tests whether AI agents can complete real clinical tasks end-to-end inside a functioning electronic health record environment, and HealthAdminBench, which explores administration and workflow. These benchmarks tell the same story: Today’s best AI agents complete less than half the work. Agents lose track of information as they move acr

匹配字段正文

后，混元的Frontier团队发布了Terminal-RSI。这次更有意思，它能通过RSI，让一个模型超越自己原本的框架，实现自己本来不具备的能力。 就像你有一台微波炉，它的功能是给食物加热。加入Terminal-RSI之后，不仅能让微波炉原本的加热功能效果更好，你还能用它来看电视、听歌、打游戏。 以Terminal-Bench 4.0为例，这是当下最严格的Agent终端实操基准测试集之一。论文使用的基座模型，原本不具备跑这个测试集的能力，也就是0%。 通过Terminal-RSI，这个基座模型能达到19.84%的分数，和GPT-5.6 Terra、Gemini 3.8 Flash齐平。 还有一个细节也挺好玩的，论文所使用的基座模型是Qwen3.8-27B，腾讯阿里也因此在RSI上达成了“合作”。 论文自己也承认，这套RSI只在一个27B的模型上跑过，没有在更大的模型上验证，也只有三次评测、没有独立训练种子，所以它放到Hy4这种大参数量模型上会怎样，还能否实现RSI，目前都还没有答案。 Terminal-RSI是个啥？ Terminal-RSI所训练的叫做终端Agent。它跟普通的Ag

匹配字段正文

offers one of the earliest looks at what an AI-dominated company of the future might look like—and spoiler alert, it includes babysitting a whole lot of bots. Recommended Video AI agents are now moving so fast that 80% of the elapsed time in Claude sessions was the result of agents waiting on a human’s input, Anthropic’s report found. “This acceleration makes human attention the bottleneck for agent output,” the report read. Instead of actively doing the work themselves, even non-supervisor empl

匹配字段正文

将增加，中位数预期增长66%。公司可以因效率提升扩张业务，但新人入门岗位可能收窄——Anthropic内部数据显示初级技术员工占比已从21%降至14%。2026-10-20 13:59AI公司员工眼中的2030年，与普通人完全不同冷眸速览 如果一位刚入职的年轻工程师打开招聘软件，他最关心的也许是：今年还有多少初级岗位？而在硅谷一家前沿AI实验室里，工程师正让Agent写代码、检查漏洞、执行一部分过去需要自己完成的任务。两个人使用的可能是相近的AI技术，对工作的感受却未必相同。 Anthropic在2026年10月16日发表《The AI Expectations Gap》，把这种差异变成了一组可以比较的数据。研究人员于当年8月访问了具有全美代表性的美国成年样本，以及每天开发和使用前沿模型的Anthropic员工。他们没有问“你喜欢AI吗”这样宽泛的问题，而是追问：AI究竟能做什么、企业会不会用、它能节省多少时间、会不会独立工作，以及被替代的人需要多久才能找到下一份职业。 结果并不是简单的“圈内人乐观、圈外人悲观”。两组人对AI今天能做什么，判断其实并不遥远；真正拉开距离的是未来四年。A

匹配字段正文

New Cohere paper finds that a team of coding agents gets the most from extra agents when they only research and plan, and cheap models can do that job for a strong lead at far lower cost. A team of coding agents beats 1 agent working alone, but costs 3 to 7 times as much. More agents can also mean duplicate work and extra code to merge. The researchers built a lead agent that hands subtasks to workers, then varied the workers' role, model and count. On ProgramBench, agents rebuild a whole progra

匹配字段正文

智东西腾讯三星英特尔360高通苹果谷歌IDx华为微软百度通用5G大模型小米iPhone 腾讯混元炼出Agent版RSI！基于Qwen3.8-27B，终端任务成功率翻倍 智东西10月19日消息，10月15日，由腾讯首席AI科学家、腾讯基础模型部负责人姚顺雨带领的腾讯混元前沿团队（Tencent Hy Frontier）联合新加坡国立大学、美国马里兰大学、佐治亚大学和印第安纳大学的研究人员，以阿里巴巴通义千问团队开源的Qwen3.8-27B为基座，提出一种面向终端智能体（Terminal Agent）的递归训练方法Terminal-RSI。 该研究旨在让终端智能体从过往的成功经验中学习，持续提升复杂任务的执行能力。为此，该研究团队提出Terminal-RSI，让Agent提炼成功操作的经验，再通过强化学习优化执行策略，在多轮训练中持续改进任务执行能力。 为检验这一方法的效果，该研究团队以Qwen3.8-27B为基座模型，依次经过Base、RSR-1、RPR-1、RSR-2和RPR-2五个阶段的训练，得到Terminal-RSI-27B，并在六项基准测试中评估其终端任务执行能力。 评估结果显

匹配字段正文

ness as likely associated with malicious, financially motivated intrusion activity. The exposed panels provide a detailed view of how threat actors choose targets, task autonomous agents, exploit vulnerabilities, and establish persistent access in compromised systems, the firm said. The report was written by Snorre Fagerland, a founding researcher at NorForge, an employee-owned company founded in 2026 with a US parent headquartered in Palo Alto and a European operating subsidiary. Because the pa

匹配字段正文

New Meta paper shows that frontier agents differ widely in how they learn from experience, and the best performer is not the most efficient learner. Most agent tests measure what a model can do once. They ignore whether it gets better by building tools, notes and skills from past attempts, and at what cost. They froze model weights and let agents play chess, Go, Hex and NetHack, rewriting their own tools and notes after each round. They then scored held-out games and defined plasticity as held-o

匹配字段正文

2️⃣ On 100 easier cases, where few witnesses lie: • 🏅 GPT6-Astra leading 12 Jev agents solve 96 • 🥈 GPT6-Astra alone solves 57 • 🥉 Jev swarm alone solves 38 In the easier cases each lie comes from a lone witness, while every true fact turns up twice. So a rule-based leader just counting clues + Jev swarm finds the truth and solves 97/100. "Elementary, my dear Watson."

匹配字段正文

o the hardware, for a specific problem. The post thanks Arjun Kar and Vinay Ramasesh for helpful discussions in the project’s early stages. 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

n McGinniss, Nicholas Carlini, and Steve Weis. The reproduction code on the page is dedicated to the public domain under CC0 1.0 Universal. 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

The paper on how a team of Gemini 4 agents beat Minecraft is worth a look. It is accessible example (partially because it is AI written) of spontaneous coordination among agents to accomplish a complicated goal & how their approach differs from humans/ https://t.co/brrmhydhWw https://t.co/cxlp4jTjqf

匹配字段正文

Grounding AI agents in biological reality: Building blocks for autonomous wet labs October 16, 2026 We are bringing AI agents to the wet-lab bench at Google Research to navigate complex biological landscapes. By grounding agents in physical lab robotics, we can run fast closed-loop molecular design cycles where agents learn how to learn, ultimately discovering single-digit nanomolar protein binders for cancer targets. Quick links VERID paper Agentic aptamer sandbox paper Epitope excision paper ×

匹配字段正文

n with opponents. The company says the approach is not limited to sport and is already being extended to everyday robotics tasks. Skild is also experimenting with larger groups of agents. Early four-agent soccer experiments have begun to produce behaviors including passing and coordination, according to the company. Future work is expected to explore collaborative manipulation and navigation as the number of interacting robots increases. Skild says: “This simulation ran for over a century with a

匹配字段正文

表作者本人，36氪平台仅提供信息存储空间服务。+10 好文章，需要你的鼓励大数据文摘特邀作者0收 藏+10评 论打开微信“扫一扫”，打开网页后点击屏幕右上角分享按钮微 博沉浸阅读返回顶部举报参与评论评论千万条，友善第一条登录后参与讨论提交评论0/1000你可能也喜欢这些文章Meta发布美国最强开源大模型抢入口、寻变现，大模型公司为何集体“不务正业”？亚马逊，Agent最严厉的父亲刚刚，Mac Studio 迎来小扎最强开源模型，苹果等来了最意外的助攻前钉钉总裁叶军，离职一年后创业了，赌大厂做不了也不屑做的「缝隙」AI 脸不够用，普通人也能 “ 卖脸 ” 赚钱了，最高 10000 元？苹果和 Meta 和好了，是 AI 逼的Linux之父也用AI写内核了，曾嘲讽AI「90%是营销」，如今用Claude Code写了8个补丁，性能暴涨3倍我们不该投资更多的「田柯宇」大数据文摘特邀作者 TA没有写简介，但内敛也是一种表达发表文章638篇 最近内容DeepMind AlphaProof Nexus登《Science》：AI自主破解多道56年未解Erdős数学难题18分钟前微软“站台”了一篇文章

匹配字段正文

Already Clean” — showing that many known cases of polysemanticity and superposition - causing tangled “knots” of weights that resist human interpretation - simply fall to further agentic analysis: Don’t take our word for it:Deedy@deedydasThis is one of the most important results in AI, period. TLDR: neural networks are not tangled knots (polysemanticity is wrong). You can study their own neurons directly, without first "unfolding" them with a second model. 90% of neurons in Qwen-32B have one jo

匹配字段正文

rol of a 6-axis robot arm (5 joints and a gripper) and a simple objective. We’re releasing results for our first eight tasks today. We find that frontier language models in coding agent harnesses have recently become capable of directly manipulating robot arms for simple tasks. Today, they are far too slow and expensive for any kind of production usage - nearly two orders of magnitude worse than human baselines. We expect this to change quickly. GPT-6 Astra places first in AA-Robotics v0.1, foll

匹配字段正文