以Antigravity的/teamwork推进科学与工程发展

内容来源:https://research.google/blog/advancing-science-engineering-with-antigravitys-teamwork-feature/
内容总结:
谷歌研究院近日发布博客文章,介绍了其Antigravity平台下/teamwork多智能体系统在数学、化学、系统软件和机器学习工程等领域取得的最新研究成果。相关研究均与人类专家合作完成,相关工具已向外部用户开放。
文章称,过去数周,谷歌及学术合作机构的研究团队利用Gemini 4系列模型和/teamwork多智能体工作流,攻克了多个悬而未决多年的开放性问题,在信息论、形式化验证、分子结构发现和内核优化等方面取得新进展,并在谷歌内部多项生产基础设施中实现性能提升。
研究团队指出,当前语言模型存在两大局限:一是单次生成结果不确定,二是长周期任务中容易丢失细节。为此,/teamwork系统通过并行探索配合专用验证器(如Lean 4证明检查器、未经修改的软件测试套件、执行模拟器等)自动筛选有效方案,并按步骤为每个专用智能体分配确切的文件和工具,由协调器向新接任者交接任务,审核智能体则在不受原始推理影响的干净上下文中独立核验结果,从而保证多日任务的可靠性。
在数学与理论计算机科学领域,谷歌与香港中文大学、南加州大学、卡内基梅隆大学等机构合作,利用/teamwork的长证明模式完整证明了信息论中2013年提出的Courtade–Kumar猜想,并以Lean 4形式化了全部证明,生成代码达5000万行。此外,多智能体系统还在自适应数据分析、量子计算、布尔矩阵乘法、高维几何、随机线性代数、动态图割稀疏化、在线逆线性优化、整数乘法、双端队列猜想及高维统计等问题上取得突破,多项成果解决了STOC、FOCS等顶级会议留下的开放问题。
在形式化数学方面,谷歌云AI团队合作开发的LEAP多智能体框架自主完成了Weil猜想约180万行Lean形式化代码,涉及28800次子智能体调用;还首次形式化了希尔伯特第十问题在一般整数环上的近期证明、三维Navier–Stokes方程有限时间爆破证明、Birkhoff猜想、三维Bochner–Riesz猜想等,并在过程中发现并修复了近期非形式化证明中的一处漏洞。此外,团队还为STOC 2027的578篇投稿提供了Deep Review反馈服务,调查显示90%的受访者认为反馈基本符合事实,超过30%的受访者表示该工具发现了实质性证明错误。
在实验科学领域,团队将30个专用智能体组织为自校正循环,结合物理信息光谱模拟器,从核磁共振数据推断分子结构,在复杂基准分子上实现95%的Top-1精确匹配和100%的Top-3准确率,接近单智能体性能的两倍。
在系统与硬件工程方面,团队利用/teamwork和Argon模型构建了低延迟内核旁路分区全局地址空间系统FarArray,其上的无锁连通分量应用可扩展至万亿边图,吞吐量提升60至90倍;InstantEmbedding将峰值内存降低至多14倍,运行时间缩短11至27倍;为图神经网络定制的TPU内核实现高达185倍的高带宽内存占用降低。团队还让爬山智能体从基础AlphaZero实现出发,在不到8天内将9x9围棋引擎的Elo提升约2600分至1885分。在大规模代码迁移方面,/teamwork将Zircon微内核剩余C++代码改写为约80万行安全Rust代码并通过全部内核单元测试,还将libgav1视频解码器改写为约17万行Rust代码,实现100%位精确一致性,且性能较原C++版本提升20%至30%。
谷歌表示,外部科学家和开发者目前可在Antigravity平台使用Gemini 4模型配合/teamwork和/boost,并已开放申请,为1万名学术研究人员提供为期六个月的最高使用额度免费访问。未来数周还将把形式化证明模式等更多专用工作流引入Antigravity,并配套发布深度教程。
中文翻译:
2026年11月13日
Vahab Mirrokni,Google Fellow 暨 Google Research 副总裁
我们分享 Antigravity /teamwork 多智能体系统的最新成果,该系统增强了人机协作能力,用于解决数学、化学、系统软件和机器学习工程领域的开放性问题。所有工作均与人类专家合作完成,我们正将这些工具对外开放。
过去数周,Google 各团队及学术合作机构的研究人员攻克了数学、科学和复杂系统工程领域悬而未决多年的开放性问题。借助 Gemini 4 Argon、Gemini 4 Carbon、Gemini 4 Hydrogen 以及 Antigravity /teamwork 等多智能体工作流(基于多种 Gemini 模型构建),这些合作在信息论、形式化验证、分子结构发现和内核优化等方面取得了新突破,同时为 Google 带来了多项生产基础设施方面的收益。我们正在 Antigravity /teamwork(连同 /boost 及其他智能体)中开放这些智能体工具,供研究人员和开发者应用于自身工作。
前沿研究与系统工程需要浏览海量文献、构建定制计算流水线、对照物理测量进行验证,并逐步核验每一个环节。进展来自一个交互式发现循环:研究人员提出问题并定义理论或物理约束,而长时程智能体则编写定制求解器、探索大规模搜索空间、测试候选设计并核查长链条逻辑。
当前语言模型的两个局限促成了 /teamwork 的设计。第一,模型生成具有不确定性,单次尝试与找到正确答案之间存在差距。该系统通过将并行探索与专用证伪器(如 Lean 4 证明检查器、未经修改的软件测试套件、执行模拟器、网络搜索或硬件计数器)配对,自动验证哪些候选方案有效,从而弥合这一差距。第二,AI 模型在多天任务中常常丢失细节,因为注意力在不断增长的上下文窗口中分散。为防止这一问题,/teamwork 通过构造方式划分上下文:每个专用智能体仅接收特定步骤所需的确切文件和工具。协调者将任务交接给全新的后继者,而评论者和审计者则在干净的上下文中对结果提出质疑和验证,不携带原始产出者的推理过程。
Teamwork 将这些工作流表达为声明式、自我改进的自然语言模式,并将每个用户提示动态路由到最适合该任务的模式。这些模式涵盖从分布式编码和迭代问题求解(作为 /boost 独立打包,即从先前轨迹中精炼而来的顺序求解-验证循环)到专业化研究工作流(如长证明和文档审查),所有这些现已在 Antigravity 中可用。这使团队能够根据过往运行记录设计定制规范,在不同执行框架模式间路由,并动态改进智能体执行框架。虽然其中许多功能已在 Antigravity 中可用,我们将在未来数周提供更多更新,向 Antigravity 的外部用户开放全部功能。继此前关于 Antigravity /teamwork 的博文之后,我们在此展示由新版 /teamwork 和 Gemini 4 模型赋能的开放性挑战新成果。
在数学和理论计算机科学领域,研究人员借助 Gemini 4 和 /teamwork 执行框架,合作完成大规模证明,并直接从开放问题陈述生成完整算法和证明。我们展示若干示例。
来自 Google、香港中文大学、南加州大学和卡内基梅隆大学的研究人员使用 /teamwork 的长证明模式(在 Antigravity 中实现 Stellar Colosseum),完整地证明了信息论中 2013 年的 Courtade–Kumar 猜想。该结果证实:当随机比特串通过噪声信道时,将输入压缩为单个比特的任何方案所保留的信息量,都不会超过“独裁函数”——即仅复制单个输入比特的函数。Gemini 首先发现了香港中文大学团队此前论文中广义猜想的反例。随后香港中文大学团队提出了一个新的候选函数,将该猜想归约为一个 14 变量优化问题,并给出了一个关键的分析下界。Gemini 4 Argon 及更早的 Gemini 模型随后划分参数空间,发展了大部分新的分析下界和区间算术证书。它们还在无人工输入的情况下将证明从平衡函数扩展到非平衡函数。团队在 Lean 4 中形式化了整个证明,生成了 5000 万行代码。
研究人员还直接向 Gemini 4 及 Stellar Colosseum 和 /teamwork 等多智能体执行框架提出开放性问题,获得了端到端算法和证明(通常附带 Lean 4 形式化),并由作者验证、简化和编辑后发表。在自适应数据分析中,运行于 Gemini 4 Argon 的多智能体执行框架独立证明并经 Lean 验证了近乎最优的样本界,适用于根据先前答案自适应选择的统计查询,填补了自 2014 年以来悬而未决的空白。这些界产生了首个同时支持插入和删除的数据流近空间最优算法,且在面对自适应对手时保持准确。在量子计算中,一项新构造解决了 2014 年的一个开放问题,首次找到需要严格超过 n/2 次量子查询(在常数误差下)的 n 位布尔函数,同时精确确定了对称函数上的查询开销。
多项结果突破了处理与压缩大型矩阵和图长期存在的壁垒。截至 STOC 2024 的五十年间,布尔矩阵乘法和三角形检测的组合算法(避免快速矩阵乘法的代数消去)仅比立方时间节省了对数因子。一种新的确定性算法解决了 STOC 2024 的核心开放问题,实现了其相对于立方时间的目标加速。在高维几何中,一个新下界解决了一个关于稀疏网络的开放问题,证明在任意常数因子内近似 n 个点之间所有成对距离,可能需要多项式级别多于 n 条边,即使允许额外的 Steiner 点和任意多跳。在随机线性代数中,堆叠稀疏草图解决了 2025 年的一个开放问题,打破了将矩阵压缩至线性维度而不使 r 维子空间任何方向坍缩的经典稀疏性壁垒,每行仅需 √log r 个非零元(比保留所有长度所需的 log r 平方级更少),并几乎匹配一个新下界。对于动态边权插入和删除下的加权图,Stellar Colosseum 解决了 FOCS 2022 遗留的开放问题,开发了首个单遍流式割稀疏化算法,同时实现近最优的内存、更新时间和重建时间。
我们的成果还涵盖优化、算术、数据结构和统计学。在优化方面,解决了一个关于在线逆线性优化(从随时间做出的决策中推断专家的隐藏目标)的开放问题,该方法给出了一种确定性算法,在实现最优遗憾的同时将运行时间从维度上的指数级降至多项式时间。对于 n 位整数乘法,多智能体攻关将理论图灵机界简化至 n log n 以下,另一项独立工作则在单个二维网格上给出了新的 O(n log n) 算法,其 C 实现在十亿比特以内与 GNU 多精度算术库(GMP)逐位一致。在数据结构方面,另一项结果在 Tarjan 1985 年的双端队列猜想(自调整伸展树能否在常数摊还时间内处理两端的插入和删除)上取得重大进展,证明了每次操作开销的新上界,同时优于所有有限迭代的逆 Ackermann 函数,并已在 Lean 4 中验证。最后,在高维统计中,该方法证明了近似留一交叉验证(避免从头重新拟合模型)对于弹性网等非光滑稀疏惩罚仍然准确。
手工检查计算机辅助的手稿给同行评审带来了沉重负担,微妙的缺陷可能隐藏在冗长的计算中。在 Lean 4 中形式化数学将证明转化为计算机内核逐行检查的代码。一旦人类数学家确认形式化定理陈述与原始问题一致,证明只有在每一步逻辑都成立时才能编译通过。
数学中许多最深层的开放问题关乎物理波如何传播、动力系统如何演化,以及多项式方程是否有整数解。我们的多智能体框架 LEAP 与 Google Cloud AI 的同事合作开发,配合 Argon,直接从非形式化证明出发,在这些核心领域取得了进展。
值得注意的是,/teamwork-LEAP 完全自主地端到端产出并经机器检查的形式化了 Weil 猜想——这一关于有限域上多项式方程解计数的里程碑成果,最终由 Pierre Deligne 于 1974 年完成证明。该形式化涵盖约 180 万行 Lean 代码,经由 28,800 次子智能体调用生成。该系统还合成了超过 919,000 行 Lean 代码,首次形式化了一般整数环上希尔伯特第十问题的最新证明。它还独立复现了三维 Navier–Stokes 方程(光滑外力情形)有限时间爆破最新证明的端到端形式化——这是七大千禧年大奖难题之一,已悬置约 90 年。
除此之外,它还生成了 Birkhoff 猜想的首次形式化,这是一个自 1927 年以来悬而未决的动力系统问题。此外,它首次形式化了三维 Bochner–Riesz 猜想——调和分析中悬置 50 多年的核心问题——并在此过程中发现并修复了最近发布的一份非形式化证明中的漏洞。其他经机器检查的突破包括:解决了三维及以上维度无穷调和函数的 C¹ 正则性(悬置约 20 年的开放问题);形式化了 Killip 和 Vişan 关于 KdV 水波方程在 H⁻¹ 中适定性的定理;以及验证了加州大学洛杉矶分校团队对 40 年历史的 Jamison 毛毛虫猜想、Ağırseven-Ollis 猜想和 Erdős 问题 #1063 改进上界的非形式化证明。
最后,我们为另一个计算机科学会议的作者提供了深度评审(Deep Review):在 STOC 2027 之前,578 篇投稿(超过全部投稿的三分之一)的作者选择接收对其草稿的深度评审反馈。在一项正在进行的作者调查中(N = 70),90% 的受访者认为反馈大部分或几乎全部符合事实,90% 表示会再次使用该工具。超过 30% 表示深度评审发现了实质性证明漏洞或数学上重要的笔误,14% 表示发现了需要超过一小时修复的技术缺陷。受访作者分享道:“这类工具可以帮助更快地产出更干净、更精确的研究”,深度评审“为科学发现提供了安心感”。我们每周向用户免费提供两次深度评审调用。深度评审也实现为 /teamwork 的文档审查模式,并已在 Antigravity 中可用。更多关于深度评审的信息请参阅我们近期的博文。
除形式化证明外,将专用生成器智能体与独立验证器智能体配对,可在实验实验室科学中实现发现工作流。
从实验核磁共振数据确定分子结构是化学中最难的谜题之一。例如,在 HIV 蛋白酶抑制剂中,一个错位的键就能将无活性的相似物与药物区分开来。通过 /teamwork 将 30 个专用智能体组织成自校正循环,并配合物理信息光谱模拟器,我们在复杂基准分子上实现了 95% 的 top-1 精确匹配和 100% 的 top-3 准确率,几乎是单智能体性能的两倍。这不仅教会了 Gemini 对结构进行推理,还为正在进行的三维蛋白质和复杂结构动力学研究奠定了基础。
除这一示例外,Gemini 4 博文中报道的包括核聚变和量子算法应用在内的科学问题也受益于 /teamwork 和 Argon 模型。
我们还应用了类似的多轮审计工作流来构建实证基准。从五项人工指定的验收标准(如数据匿名化)出发,29 个运行于 Argon 的智能体在 11.5 小时内将 13 张公开表格转化为一个 50 任务的因果推断基准。在这些任务中,模型必须判断因果效应是否可识别,并对照已知答案进行估计。经过初步检查后,独立的评审者和审计者智能体捕捉捷径、标记措辞暗示、否决幻觉脚本,将每一项转化为永久测试。自动化这一高度人工化的任务生成过程,提供了帮助 Gemini 回答反事实“如果……会怎样”问题所需的难题。
当与客观评估器(如网络基准、游戏引擎、未经修改的测试套件和硬件计时器)配对时,/teamwork 攻关活动可以执行多天的软件和硬件工程项目。
使用 /teamwork 和 Argon,我们构建了 FarArray——一个基于 SNAP 微内核的低延迟内核旁路分区全局地址空间。基于 FarArray 构建的应用包括无锁连通分量算法,可扩展至万亿边图,相比基于 TCP 的键值存储实现 60 倍至 90 倍的吞吐量提升和 40 倍至 180 倍的加速,同时内存占用降低 6 倍。类似地,InstantEmbedding(个性化 PageRank)相比加速的 MapReduce 实现,峰值 DRAM 降低最多 14 倍,运行时间缩短 11 倍至 27 倍。团队还使用 /teamwork 和 Argon 为图神经网络创建了定制 TPU 内核,实现高达 185 倍的高带宽内存降低以及前向和反向传播 8 倍至 15 倍的加速。更多细节请参阅我们近期的 Antigravity 博文。
为测试长时程自主优化,Argon 上的爬山团队从基础 AlphaZero 实现出发,从零训练了一个 9x9 围棋引擎。人类研究人员设定芯片分配和评估设置,而执行框架运行自主攻关以实现更快的自对弈(达到 2,300 倍加速)和更快的 Elo 读数,严格通过锁定记录上的统计检验来晋升冠军。在不到八天内,该引擎提升了约 2,600 Elo,达到 1,885 Elo,位于每步 16 次和 32 次访问的 KataGo 之间。
大型代码迁移在智能体用桩替代依赖或修改测试以使损坏的移植通过时就会失败。一个新的 /teamwork 重写模式(即将在 Antigravity 中提供)通过仅在依赖编译后才调度每个模块、将每个叶模块链接到原始未经修改的测试套件来解决这一问题。在 Zircon 微内核初始 25% 移植的基础上,为期两周的 Argon 攻关重写了剩余的 C++ 代码,新增约 80 万行安全 Rust 代码,通过所有内核单元测试并在 arm64 硬件上启动。/teamwork 还将 Google 的 AV1 视频解码器 libgav1 从 C++ 重写为约 17 万行 Rust,在广泛的合规测试套件上实现 100% 位精确一致。最初,移植版本比原始 C++ 代码慢约 4 倍。然而,在给予 /teamwork 访问性能分析工具并赋予在样本视频片段上最大化速度的任务后,Rust 移植版在 x86_64 和 ARM 上均比原始 C++ 和手工优化汇编代码快 20-30%。
对于日常编码任务,Antigravity /boost 命令打包了 /teamwork 的迭代编码模式。一项长期运行的 /teamwork 攻关分析了过往运行轨迹,发现紧耦合代码最好通过顺序实现和修订阶段而非并行拆分来处理,从而改进了 /boost 在编码任务上的表现。这是一个执行框架为令牌效率优化另一个执行框架的示例。
对于机器学习系统,定制 JAX 和 Pallas TPU 内核的完整硬件评估需要 7 至 32 分钟,限制了先前优化智能体所能探索的候选数量。使用 Argon 上 /teamwork 的 DuoDiscover 模式,四个并行智能体识别性能瓶颈并构建了快速、23 秒的代理评估器,称为“标尺”。这些标尺产生的候选排序通过 Kendall 秩相关与周期性完整硬件运行持续验证。DuoDiscover 在 17 个 JAXBench 内核上相比参考实现实现 4.09 倍几何平均加速,在 7 个手工优化的前沿开放权重模型 vLLM TPU 推理 Pallas 内核上实现 2.99 倍几何平均加速。
我们还将 /teamwork 和 /boost 应用于 Google 内许多生产基础设施和硬件建模问题。示例包括:
外部科学家和开发者即日起可在 Antigravity 和深度评审中使用 Gemini 4 模型配合 /teamwork 和 /boost。我们还开放了申请通道,为 10,000 名学术研究人员提供六个月免费使用我们最高用量层级的机会。即将到来的更新将把我们的形式化证明模式和更多专业工作流引入 Antigravity。这些功能将伴随未来数周 Antigravity 博文上的深入教程,确保开发者和研究人员拥有生成、检查和迭代精炼自定义多智能体团队所需的资源。我们很高兴将我们的工具(以及我们的模型)开放给所有人。
Antigravity 的 /teamwork 功能是 Google 全公司努力的成果,凝聚了许多贡献者和持续探索。贡献者包括:Yuan Deng
多位为研究论文做出贡献的学术合作者:Zijie Chen、Amin Gohari、Chandra Nair(香港中文大学),Songbin Liu、Arian Maleki、Chen Shang、Hanbo Yang、Wenjin Zhang(哥伦比亚大学),Zihan Hao、Qipeng Liu、Jianan Shao(加州大学圣地亚哥分校),Lorenzo Beretta(IBM 剑桥),Simone Bombari(杨百翰大学 & Flatiron 研究所),Mahbod Majid、Margalit Glasgow(麻省理工学院),Raphael Arkady Meyer(加州大学伯克利分校、ICSI & 劳伦斯伯克利国家实验室),Kijun Shin(卡内基梅隆大学),Huacheng Yu、Elena Gribelyuk(普林斯顿大学),Xin Lyu(加州大学伯克利分校/字节跳动),Samson Zhou(德克萨斯农工大学),Erik Waingarten(宾夕法尼亚大学),以及 Sorrachai Yingchareonthawornchai(ETH-ITS)。
英文来源:
November 13, 2026
Vahab Mirrokni, Google Fellow and VP, Google Research
We share recent results from Antigravity’s /teamwork multiagent systems, which enhance human-AI collaborations to solve open problems across mathematics, chemistry, systems software, and ML engineering. All of this work is done in collaboration with human experts, and we are making these tools available externally.
Over the past several weeks, teams across Google and academic partner institutions have tackled open problems in math, science, and complex systems engineering that had remained open for years. Using Gemini 4 Argon, Gemini 4 Carbon, Gemini 4 Hydrogen and multi-agent workflows such as Antigravity’s /teamwork (building on various Gemini models), these collaborations have produced new innovations across information theory, formal verifications, molecular structure discoveries, and kernel optimization, along with several production infrastructure gains at Google. We are making these agentic tools available in Antigravity’s /teamwork (alongside /boost and other agents) so that researchers and developers can apply them to their own work.
Frontier research and systems engineering require navigating vast literature, building custom computational pipelines, testing against physical measurements, and verifying every step. Progress comes from an interactive discovery loop where researchers frame questions and define the theoretical or physical constraints, while long-horizon agents write custom solvers, explore large search spaces, test candidate designs, and check long chains of logic.
Two limits of current language models motivate /teamwork's design. First, model generation can be unpredictable, leaving a gap between a single attempt and finding the correct answer. The system bridges this gap by pairing parallel exploration with dedicated falsifiers (e.g., the Lean 4 proof checker, unmodified software test suites, execution simulators, web search, or hardware counters) to automatically verify which candidate solutions work.
Second, AI models often lose track of details over multi-day tasks because attention disperses across growing context windows. To prevent this, /teamwork partitions context by construction: each specialized agent receives only the exact files and tools it needs for a specific step. Coordinators hand off tasks to fresh successors, while critics and auditors challenge and verify results from clean contexts without the original producer's reasoning.
Teamwork expresses these workflows as declarative self-improving natural-language patterns and dynamically routes each user prompt to the pattern best suited to the task. These patterns range from distributed coding and iterative problem-solving (packaged standalone as /boost, a sequential solve-and-verify loop refined from prior trajectories) to specialized research workflows like long proof and document review, all of which are available in Antigravity today. This allows teams to design custom specifications from past run transcripts routing across different harness patterns and dynamically improving agentic harness. While many of these features are available in Antigravity, we will provide more updates in the coming weeks providing all the features to our external users on Antigravity. Following our previous posts on Antigravity /teamwork, we provide new results to open challenges empowered by new variants of /teamwork and Gemini 4 models.
Across mathematics and theoretical computer science, researchers paired with Gemini 4 and /teamwork harnesses to collaborate on large-scale proofs and generate complete algorithms and proofs directly from open problem statements. We present several examples.
Researchers from Google, CUHK, USC, and CMU used /teamwork's Long Proof pattern (implementing Stellar Colosseum in Antigravity) to prove information theory's 2013 Courtade–Kumar conjecture in full generality. The result confirms that when a random bit string passes through a noisy channel, no compression of the input to a single bit preserves more information than a "dictator function", which just copies a single input bit. Gemini first found counterexamples to a generalized conjecture from the CUHK team's earlier paper. The CUHK team then proposed a new candidate function that reduced the conjecture to a 14-variable optimization problem, along with a key analytical lower bound. Gemini 4 Argon and earlier Gemini models then partitioned the parameter space and developed most of the new analytical bounds and interval-arithmetic certificates. They also extended the proof from balanced to unbalanced functions without human input. The team formalized the entire proof in Lean 4, producing 50 million lines of code.
Researchers also posed open problems directly to Gemini 4 and to multi-agent harnesses such as Stellar Colosseum and /teamwork, receiving end-to-end algorithms and proofs (often with Lean 4 formalizations) that the authors verified, simplified, and edited for publication. In adaptive data analysis, a multi-agent harness running on Gemini 4 Argon independently proved and Lean-verified near-optimal sample bounds for statistical queries chosen adaptively from earlier answers, closing a gap open since 2014. These bounds yield the first nearly space-optimal algorithm for data streams with both insertions and deletions that stays accurate against adaptive adversaries. In quantum computing, a new construction settled a 2014 open question by finding the first n-bit Boolean functions needing strictly more than n/2 quantum queries at constant error, while pinning down the exact query overhead across symmetric functions.
Several results break long standing barriers in processing and compressing large matrices and graphs. For fifty years until STOC 2024, combinatorial algorithms for Boolean matrix multiplication and triangle detection (avoiding the algebraic cancellations of fast matrix multiplication) saved only logarithmic factors over cubic time. A new deterministic algorithm resolves STOC 2024's main open question, achieving its target speedup over cubic time. In high-dimensional geometry, a new lower bound resolves an open problem on sparse networks, proving that approximating all pairwise distances among n points within any constant factor can require polynomially more than n edges, even with extra Steiner points and arbitrarily many hops. In randomized linear algebra, stacked sparse sketches resolve a 2025 open question by breaking the classical sparsity barrier for compressing a matrix to linear dimension without collapsing any direction of an r-dimensional subspace, needing only √log r nonzeros per row (quadratically fewer than the log r required to preserve all lengths) and nearly matching a new lower bound. For weighted graphs under dynamic edge-weight insertions and deletions, Stellar Colosseum resolved an open question left by FOCS 2022, developing the first single-pass streaming algorithm for cut sparsification that achieves near-optimal memory, update time, and reconstruction time simultaneously.
Our results also span optimization, arithmetic, data structures, and statistics. In optimization, resolving an open question on online inverse linear optimization (inferring an expert's hidden objective from decisions over time), the approach gave a deterministic algorithm achieving optimal regret while cutting runtime from exponential in the dimension to polynomial time. For n-bit integer multiplication, multi-agent campaigns simplified the theoretical Turing-machine bound below n log n, while a separate effort gave a new O(n log n) algorithm on a single two-dimensional grid whose C implementation matches the GNU Multiple Precision Arithmetic Library (GMP) bit-for-bit up to a billion bits. In data structures, another result made significant progress on Tarjan's 1985 Deque Conjecture (whether self-adjusting splay trees handle insertions and deletions at both ends in constant amortized time), proving a new upper bound on the per-operation overhead that simultaneously beats every finitely iterated inverse-Ackermann function, verified in Lean 4. Finally, in high-dimensional statistics, the approach proved that approximate leave-one-out cross-validation, which avoids refitting the model from scratch, remains accurate for non-smooth sparse penalties like elastic net.
Checking a computer-assisted manuscript by hand places a heavy burden on peer reviewers, and subtle flaws can hide inside long calculations. Formalizing mathematics in Lean 4 turns a proof into code that a computer kernel checks line by line. Once human mathematicians confirm that the formal theorem statement matches the original problem, the proof only compiles if every logical step holds.
Many of the deepest open problems in mathematics govern how physical waves propagate, how dynamical systems evolve, and whether polynomial equations have integer solutions. Working directly from informal proofs, our multi-agent framework, LEAP, developed with colleagues in Google Cloud AI and paired with Argon, has made progress in these core areas.
Notably, /teamwork-LEAP fully autonomously produced an end-to-end, machine-checked formalization of the Weil conjectures, the landmark results on counting solutions of polynomial equations over finite fields that culminated in Pierre Deligne's 1974 proof. This formalization spans about 1.8 million lines of Lean, produced across 28,800 subagent invocations. The system also synthesized over 919,000 lines of Lean to produce the first formalization of a recent proof of Hilbert's tenth problem over general rings of integers. It also independently reproduces an end-to-end formalization of a recent proof of finite-time blowup for the 3D Navier–Stokes equations with smooth forcing, part of one of the seven Millennium Prize Problems and open for about 90 years.
Beyond these, it also generated the first formalization of the Birkhoff conjecture, a dynamical-systems problem open since 1927. Moreover, it formalized for the first time the 3D Bochner–Riesz conjecture, a central problem in harmonic analysis open for more than 50 years, finding and fixing a gap in a recently released informal proof along the way. Additional machine-checked breakthroughs include resolving the C¹ regularity of infinity-harmonic functions in three and higher dimensions (an open question for about 20 years), formalizing Killip and Vişan's theorem that the KdV water-wave equation is well-posed in H⁻¹ and verifying a UCLA group's informal proofs of the 40-year-old Jamison Caterpillar Conjecture, the Ağırseven-Ollis conjecture, and an improved upper bound on Erdős Problem #1063.
Finally, we provided Deep Review for authors of another CS conference: ahead of STOC 2027, authors of 578 submissions (over a third of all submissions) opted in to receive Deep Review feedback on their drafts. In an ongoing author survey (N = 70), 90% of respondents rated the feedback mostly or almost all factual and 90% said they would use the tool again. Over 30% said Deep Review caught substantive proof bugs or mathematically significant typos, and 14% said it found technical gaps that took more than an hour to fix. Surveyed authors shared that “Tools of this kind can help produce cleaner and more precise research, faster” and that DeepReview provides peace of mind on scientific discovery”. We provide users two weekly calls to Deep Review for free. Deep Review is also implemented as /teamwork's document review pattern and available in Antigravity. Read more about DeepReview in our recent blog post.
Beyond formal proofs, pairing specialized generator agents with independent verifier agents enables discovery workflows in experimental laboratory science.
Determining a molecule’s structure from experimental nuclear magnetic resonance data is one of chemistry’s hardest puzzles. For example, in HIV protease inhibitors, one misplaced bond separates an inactive look-alike from a drug. By organizing 30 specialized agents into a self-correcting loop with /teamwork and a physics-informed spectral simulator, we achieved 95% top-1 exact match and 100% top-3 accuracy on complex benchmark molecules, nearly doubling single-agent performance. Beyond teaching Gemini to reason over structure, this lays the groundwork for ongoing works on 3D protein and complex structures dynamics.
Beyond this example, science problems including applications in nuclear fusion and quantum algorithms reported in the Gemini 4 blog post benefited from /teamwork and the Argon model.
We also applied a similar multi-round auditing workflow to build empirical benchmarks. Starting from five human-specified acceptance criteria like data anonymization, 29 agents running on Argon over 11.5 hours turned 13 public tables into a 50-task causal inference benchmark. In these tasks, models must determine whether a causal effect can be identified and estimate it against a known answer. After initial checks, separate reviewer and auditor agents caught shortcuts, flagged wording hints, and vetoed hallucinated scripts, turning each into a permanent test. Automating this heavily manual task generation supplies the hard problems needed to help Gemini answer counterfactual "what-if" questions.
When paired with objective evaluators such as network benchmarks, game engines, unmodified test suites, and hardware timers, /teamwork campaigns can execute multi-day software and hardware engineering projects.
Using /teamwork and Argon, we built FarArray, a low-latency kernel-bypass Partitioned Global Address Space based on the SNAP microkernel. Applications built on FarArray include lock-free Connected Components, which scale to 1-trillion-edge graphs with 60x to 90x throughput gains and 40x to 180x speedups at 6x lower RAM over TCP key-value stores. Similarly, InstantEmbedding (Personalized PageRank) reduced peak DRAM by up to 14x and runtime by 11x to 27x over accelerated MapReduce implementations. The team also used /teamwork and Argon to create custom TPU kernels for Graph Neural Networks, delivering up to a 185x High Bandwidth Memory reduction and 8x to 15x speedups in forward and backward passes. Further details are available in our recent Antigravity blog post.
To test autonomous optimization over long horizons, hill-climbing teams on Argon trained a 9x9 Go engine from scratch starting from a basic AlphaZero implementation. Human researchers set the chip allocation and evaluation setup, while the harness ran autonomous campaigns for faster self-play (achieving a 2,300x speedup) and faster Elo readings, promoting champions strictly via statistical tests on locked records. In under eight days, the engine gained roughly 2,600 Elo, reaching 1,885 Elo, placing it between KataGo at 16 and 32 visits per move.
Large code migrations fail when agents stub out dependencies or alter tests to make a broken port pass. A new /teamwork rewrite pattern, soon to be available in Antigravity, solved this by scheduling each module only after its dependencies compile, linking every leaf module against the original, unmodified test suite. Building on the initial 25% port of the Zircon microkernel, a two-week Argon campaign rewrote the remaining C++ code, adding about 800,000 lines of safe Rust that pass all kernel unit tests and boot on arm64 hardware. /teamwork also rewrote libgav1, Google's AV1 video decoder, from C++ into around 170k lines Rust, achieving 100% bit-exact parity across a broad conformance test suite. Initially, the port was around 4x slower than the original C++ code. However, after giving /teamwork access to profiling tools and a mandate to maximize speed on sample video clips, the Rust port became 20-30% faster than the original C++ and hand-optimized assembly code on both x86_64 and ARM.
For day-to-day coding tasks, the Antigravity /boost command packages /teamwork's iterative coding pattern. A long-running /teamwork campaign analyzed past run trajectories and found that tightly coupled code is best handled through sequential implementation and revision stages rather than parallel splits, improving /boost on coding tasks. This is an example of a harness optimizing another harness for token efficiency.
For machine learning systems, full hardware evaluations of custom JAX and Pallas TPU kernels take 7 to 32 minutes, limiting how many candidates prior optimization agents can explore. Using /teamwork's DuoDiscover pattern on Argon, four parallel agents identified performance bottlenecks and constructed fast, 23-second proxy evaluators called yardsticks. These yardsticks produce candidate orderings that are continuously validated against periodic full hardware runs using Kendall's rank correlation. DuoDiscover achieves a 4.09x geometric-mean speedup over reference implementations across 17 JAXBench kernels and a 2.99x geometric-mean speedup across 7 hand-optimized vLLM TPU-Inference Pallas kernels for frontier open-weight models.
We have also applied /teamwork and /boost to many production infrastructure and hardware modeling problems across Google. Examples include:
External scientists and developers can use Gemini 4 models with /teamwork and /boost in Antigravity and Deep Review today. We have also opened applications giving 10,000 academic researchers six months of free access to our highest usage tier. Upcoming updates will bring our formal proof pattern and additional specialized workflows into Antigravity. These features will be accompanied by deep-dive tutorials on the Antigravity blog in the coming weeks, ensuring that developers and researchers have the resources needed to generate, inspect, and iteratively refine custom multi-agent teams. We are excited to make our tools (in addition to our models) available for everyone.
Antigravity’s /teamwork feature is a result of a Google-wide effort with many contributors and ongoing explorations. Contributions include: Yuan Deng
Several academic collaborators contributing to the research papers: Zijie Chen, Amin Gohari, Chandra Nair (CUHK), Songbin Liu, Arian Maleki, Chen Shang, Hanbo Yang, Wenjin Zhang (Columbia University), Zihan Hao, Qipeng Liu, Jianan Shao (UCSD), Lorenzo Beretta (IBM Cambridge), Simone Bombari (BYU & Flatiron Institute), Mahbod Majid, Margalit Glasgow (MIT), Raphael Arkady Meyer (UC Berkeley, ICSI, & LBNL), Kijun Shin (CMU), Huacheng Yu, Elena Gribelyuk (Princeton University), Xin Lyu (UCB/ ByteDance), Samson Zhou (Texas A&M), Erik Waingarten (UPenn), and Sorrachai Yingchareonthawornchai (ETH-ITS).
文章标题:以Antigravity的/teamwork推进科学与工程发展
文章链接:https://news.qimuai.cn/?post=5561
本站文章均为原创,未经授权请勿用于任何商业用途