跳到正文

内容类型

论文研究 最新动态

值得读的 AI 论文与研究成果:架构创新、训练方法、能力测量与理论进展。

117 条精选近 30 天 95 条共收录 610 条

更新

精选归档 · 第 6 页

10月17日周六第 101–117 条
  1. Google DeepMind78

    Google DeepMind 转发介绍,四个 Gemini 4 Argon 智能体在同一个 Minecraft 世界中从零开始,用 27 小时击败末影龙。它们只在沙盒中用代码和文本操作,共编写 2,783 个程序,通过聊天分配任务、共享笔记并互相调试错误,全程没有人类帮助。

    引用Maxence Faldor@maxencefaldor

    Four Gemini 4 Argon agents, one shared Minecraft world, one goal: defeat the Ender Dragon. 27 hours and 2,783 self-written programs later, the dragon fell. https://x.com/i/article/2111007444981645312

    推荐理由:四个 Gemini 4 Argon 智能体在共享沙盒中自行分工、写码与调试,展示了多智能体协作完成长程任务的一种路径。

10月16日周五
  1. Microsoft Research60

    微软研究院发布 Magnetar:基于标准哈希函数的通用零知识证明系统

    微软研究院发布零知识虚拟机 Magnetar,可直接证明用 Rust、Go 或 C 编写的程序执行过程,并输出可在约 12 毫秒内验证的简短证明。在 15 个基准程序上,它比同类最快的 zkVM 平均快 10 到 39 倍,在笔记本上每秒可证明 7.2 至 8.7 百万条 RISC-V 指令。

    推荐理由:微软研究院梳理十余年零知识证明研究脉络,说明 Magnetar 如何在通用性、速度与密码学假设之间同时取得进展。

10月13日周二
  1. Demis Hassabis74

    Google 与 BIDMC_Medicine 合作的 AMIE 研究在《柳叶刀》主刊发表,这是面向患者的对话式诊断首次在真实诊所开展前瞻性研究。在真实急诊患者中测试显示,临床医生在 75% 的病例中认为 AMIE 生成的摘要有助于就诊准备,超过半数病例中影响了其诊疗思路;AMIE 的鉴别诊断与医生最终诊断的一致率为 90%。Demis Hassabis 转发该消息,称这是研究从实验室走向真实临床应用的里程碑。

    引用Sundar Pichai@sundarpichai

    Today @TheLancet published our AMIE study with @BIDMC_Medicine, the first prospective study of patient-facing conversational diagnostics in a real clinic. AMIE is a research system people can chat with before doctors appointments. Testing AMIE outside the lab with real urgent care patients showed how technology can help gather medical histories safely. Clinicians reported that the summaries helped them prepare for visits in 75% of cases and influenced their approach to care in more than half. AMIE’s differential diagnoses also matched the doctors’ final diagnoses 90% of the time giving clinicians more time to focus on patient care. It’s Google’s first-ever paper in The Lancet’s main journal!

    推荐理由:AMIE 在真实急诊场景的前瞻性研究给出了临床医生采纳率与诊断一致率的具体数字,可作为判断对话式诊断落地程度的参照。

10月7日周三
10月2日周五
9月30日周三
  1. Google DeepMind66

    Google DeepMind 推出 SynthID Bio,为 AI 生成的蛋白质加可验证水印

    Google DeepMind 发布 SynthID Bio,把水印技术引入合成生物学,将不可感知的签名嵌入生物代码,在实验室测试中保持蛋白质生物功能不变。该方法针对序列微调氨基酸选择、针对 3D 结构调整原子坐标,并在 AlphaFold 3 扩散网络中微调部分权重,使预测坐标自带可检测签名。

    推荐理由:Google DeepMind 把 SynthID 水印扩展到合成生物学,读者可了解其在蛋白质设计与结构预测中的嵌入方式及生物安全用途。

9月29日周二
  1. Microsoft Research62

    微软研究院推出面向生物学复杂性的 AI 研究系统 Quine

    微软研究院推出 Quine,一个由跨模态生物学世界模型和连接模型、科学工具、文献与研究者的交互框架组成的 AI 研究系统。该系统与 Broad Institute 合作用于胰腺导管腺癌研究,从数千个化合物中筛选并优先排序候选分子,湿实验显示排名最高的化合物产生了最大的经典态到基底态细胞状态转变,整个筛选到候选验证过程仅用一个周末。

    推荐理由:微软研究院公开 Quine 的生物学世界模型与实验编排框架,并给出胰腺癌化合物筛选的湿实验验证结果。

9月26日周六
9月17日周四
8月22日周六
7月15日周三
7月1日周三
  1. Jim Fan71

    Jim Fan 团队发布 Physical AutoResearch 系列第二项工作 ASPIRE,让编码智能体观察仿真与真实机器人的多模态感知轨迹,对控制程序做演化搜索,并把最优经验蒸馏进不断扩张的技能库。作者称这种持续学习以技能精炼替代梯度下降,跨仿真到真实和跨本体迁移最多可减少约 10x 的迁移学习 token,并计划开源整套技术栈。

    引用Jim Fan@DrJimFan

    Today, we give robots a /skills library that self-evolves and compounds indefinitely! Introducing ASPIRE: a robot solving its 100th task is no longer as clueless as solving its first. Coding agents observe multimodal sensory traces from simulation and real robots, launch an evolutionary search over control programs, and distill the best know-how into an ever-expanding library. ASPIRE is a new type of continual learning: "training" is skill refinement instead of gradient descent. "Trained model" is a repo of sensorimotor skills instead of floating weights. “Distributed training” is a panel of agents each practicing a different skill instead of sharded minibatches. Here's the beauty: ASPIRE gives the tired terms "sim2real transfer" and "cross-embodiment transfer" a whole new meaning. Bridging the sim-to-real gap is notoriously brutal. An end-to-end policy has to swallow both the visual shift (sim looks toyish next to a real camera) and the subtle contact physics it never quite gets right. ASPIRE sidesteps the mess, because it doesn't ship pixels or weights across the gap, but ships the know-how. The robot still has to practice in the real world, not zero-shot, but it gets there way faster because it isn't rediscovering the strategy from scratch. Same for going single-arm to bimanual hardware, which usually requires new data and retraining from zero. ASPIRE achieves up to ~10x cut in "transfer learning” tokens (yes, tokens are the new unit of *training* compute ;) Check out our gallery of 150+ tasks and 90+ skills the robots taught themselves, all on the website! Kind of wild that we can ship the "learned weights" as an HTML page rather than a GGUF. We'll open-source the full stack so your own robot library starts compounding from ours! Deep dive in thread:

    推荐理由:ASPIRE 把机器人持续学习拆成技能库演化,读者可了解其与梯度下降训练在迁移成本上的差别。

6月18日周四
  1. Jim Fan80

    Jim Fan 介绍了 ENPIRE 在按下回车前必须完成的工程设计,包括两层安全机制、冻结奖励函数和系统遥测。安全上设置硬性运动学限制,机器人越界即任务失败并自动重置,同时用限扭力柔性夹爪让错误接触安全停住;奖励上先用少量成功与失败演示训练视觉分类器,再将其冻结为实时奖励函数。

    引用Jim Fan@DrJimFan

    Today, we enable AutoResearch in the physical world for the first time! Introducing ENPIRE: we give 8 Codex agents a fleet of robots, an allocation of GPUs, and generous token budget. We set them free with a simple goal: solve the task as quickly as possible, keep the robots busy but stay safe, don't waste precious compute. Make no mistake. Then humans step aside and our watch begins. The robot fleet starts to come alive: they learn to look for visual clues, reset the scene, practice novel skills, tinker with control stack, read papers online, debate, reflect, get stuck, and try again directly on the hardware. All we did is to give Codex an API to the world of atoms, and the rest is emergence. ENPIRE is able to solve high-precision tasks like tying zip-ties, organizing fine pins, and installing GPUs all by itself. We also discovered a new type of "physical scaling": 8 robots exploring in parallel improves significantly faster than fewer ones. A part of our NVIDIA GEAR lab now self-improves tirelessly over night. We just read the reports in the morning. /goal: we all take a holiday and Jensen wouldn't even notice ;) We will be open-sourcing everything, so you can host your self-running robot lab at home too! Deep dive in the thread:

    推荐理由:作者拆解了让 8 个 Codex 智能体自主运行机器人实验室前必须搭好的安全、奖励与遥测设计。

6月17日周三
4月1日周三
2月26日周四
2月25日周三