鹭羽 发自 凹非寺 量子位 | 公众号 QbitAI 一睁眼！ 陈天桥 带队的大模型黑马MiroMind再度满血归来—— 正式发布新一代重型推理智能体： MiroThinker-1.7 和 MiroThinker-H1 。 何为重型？延续V1.5的深度推理基因，但任务更复杂、结果更精确。 眼见为实，以基准测试为例。 MiroThinker-1.7系列发布即霸榜多项深度研究任务测试，其中MiroThinker-H1刷新 SOTA ，超越Gemini-3.1-Pro、GPT-5.4-Thinking、 Claude-4.6-Opus等一众行业顶尖闭源模型： 另外开源模型MiroThinker-1.7 （235B） 和小尺寸的MiroThinker-1.7-mini （30B） 也在效率与性能之间达到了最优平衡。 换言之，针对差异化的复杂推理需求，MiroMind已经为开发者们准备好了各式精准匹配的模型方案，致力于 将算力用在刀刃上 。 BrowseComp （网页检索类大模型基准测试） ：88.2% BrowseComp-ZH （BrowseComp的中文适配版本） ：84.4% GAIA-Val-165 （GA ...

Core Viewpoint - The article discusses how players of Pokémon Go have unknowingly contributed to a massive dataset of 30 billion real-world images used for AI training, highlighting the innovative use of gaming for data collection and its implications for technology development [2][21][38]. Group 1: Data Collection and Usage - Pokémon Go has amassed 30 billion images over ten years, collected from players who engaged with the game in various weather conditions and times of day [2][21]. - The dataset includes centimeter-level precision and covers millions of high-value locations, providing a quality that traditional mapping methods cannot achieve [3][23]. - Niantic has utilized this dataset to develop a Visual Positioning System (VPS) that can replace GPS in areas where it fails, such as urban canyons and underground locations [5][27]. Group 2: Business Model and Strategy - Niantic's strategy has integrated crowdsourced mapping into its core business model since the game's inception, positioning the game as a data collection network rather than just entertainment [36][38]. - The company has attracted significant investment, reaching a peak valuation of $9 billion, partly due to this innovative approach [39]. - Niantic has even spun off its gaming business to focus on spatial AI, indicating a shift towards leveraging the collected data for future technological advancements [40]. Group 3: Real-World Applications - The VPS technology has been implemented in delivery robots, such as those from Coco Robotics, enhancing their operational efficiency by allowing them to navigate without GPS [30][32]. - These robots have completed over 500,000 deliveries, showcasing the practical application of the data collected from Pokémon Go players [33].

Core Viewpoint - The article emphasizes the ongoing AI boom and invites individuals to join the company "Quantum Bit," which focuses on tracking AI advancements and has established itself as a leading content platform in the industry [1]. Group 1: Job Opportunities - The company is hiring for three main directions: AI Industry, AI Finance, and AI Product, with positions available for both experienced professionals and fresh graduates [2][4]. - Positions are open for various levels, including editors, lead writers, and chief editors, with a focus on matching roles to individual capabilities [6]. Group 2: Job Responsibilities - **AI Industry Direction**: Responsibilities include tracking innovations in infrastructure, such as chips, AI infrastructure, and cloud computing, as well as interpreting technical reports from conferences [6][7]. - **AI Finance Direction**: Focuses on venture capital, financial reports, and capital movements within the AI industry, requiring strong analytical skills and a passion for interviews [11]. - **AI Product Direction**: Involves monitoring AI applications and hardware developments, producing in-depth evaluations of AI products, and engaging with industry experts [11]. Group 3: Benefits and Growth - Employees will have the opportunity to engage with cutting-edge AI technologies, enhance their work efficiency through new tools, and build personal influence in the AI field [6]. - The company offers competitive salaries, comprehensive benefits, and a supportive environment for professional growth, including mentorship from senior editors [6][12]. Group 4: Company Impact - By 2025, Quantum Bit aims to have over 2.4 million subscribers on WeChat and more than 7 million users across platforms, with a daily reading volume exceeding 2 million [12]. - The company is recognized as the top new media outlet in the AI and frontier technology sectors according to third-party data platforms [12].

Core Insights - The article discusses the autoresearch project initiated by Karpathy, which allows AI to autonomously conduct experiments and improve language model training efficiency by approximately 11% without human intervention [1][5] - The project evolved from a single AI conducting experiments to a distributed community of AIs collaborating on research, running over 2000 experiments in just four days [2][10] - A self-organized peer review system emerged among the AIs, indicating a significant advancement in how AI can simulate a research community [4][12] Group 1: Project Development - The autoresearch project initially consisted of 630 lines of Python code and was designed to simulate an entire research community rather than just a single PhD student [1][5] - The number of AIs involved in the project expanded from 13 to over 80 within a week, demonstrating rapid growth and collaboration [10] - A variety of roles emerged among the AIs, including experimenters, verifiers, statisticians, and meta-analysts, all without pre-assigned tasks [11][13] Group 2: Experimental Findings - A significant finding was that many claimed improvements in model performance were often just noise, with one AI discovering that seed variance accounted for approximately 0.002 BPB, which is the same magnitude as many reported improvements [25][26] - The optimal architecture identified by the AIs was unexpectedly small, consisting of 12 layers, a dimension of 512, and an aspect ratio of 40 [23] - Several well-regarded techniques failed dramatically, leading to significant performance degradation, which was documented in a shared memory system to prevent future AIs from repeating the same mistakes [27][28] Group 3: Knowledge Sharing and Optimization - The collective memory of the AIs accelerated the discovery process, allowing new AIs to build on existing knowledge rather than starting from scratch [31][32] - AIs demonstrated the ability to learn from past experiments, avoiding redundancy and enhancing the efficiency of research [9][12] - The project also highlighted the importance of adjustable parameters over fixed constants, with many improvements resulting from replacing static values with learnable parameters [21][22] Group 4: Broader Implications - The findings suggest that the most significant breakthroughs may not lie in model architecture but rather in data scheduling and pipeline management, as indicated by over 1000 hypotheses generated by meta-AIs [29][30] - The autoresearch framework has implications for future AI research, showcasing the potential for AIs to autonomously explore and optimize not just models but also scientific discovery processes [33][36] - The project has sparked interest in the broader AI community, emphasizing the need for collaboration and shared knowledge in advancing AI research [38][41]

Core Viewpoint - The article emphasizes the ongoing AI boom and invites individuals to join the company "Quantum Bit," which focuses on tracking AI advancements and has established itself as a leading content platform in the industry [1]. Group 1: Job Opportunities - The company is hiring for three main directions: AI Industry, AI Finance, and AI Product, with positions available for both experienced professionals and fresh graduates [2][4]. - Positions are full-time and based in Beijing, with various levels of roles open for application [2][4]. Group 2: Job Responsibilities - **AI Industry Direction**: Focuses on innovations in infrastructure, including chips, AI infrastructure, and cloud computing [6]. - **AI Finance Direction**: Involves tracking venture capital and financial reports in the AI sector, monitoring capital movements within the industry [6]. - **AI Product Direction**: Concentrates on the application and hardware advancements in AI, including software applications and product evaluations [6]. Group 3: Benefits and Growth Opportunities - Employees will have the chance to engage with the latest AI technologies, enhance their work efficiency through new AI tools, and build personal influence by creating original content [6]. - The company offers competitive salaries, comprehensive benefits including social insurance, meal allowances, and performance bonuses [6]. Group 4: Company Reach and Impact - As of 2025, Quantum Bit has over 2.4 million subscribers on WeChat and more than 7 million users across platforms, with a daily reading volume exceeding 2 million [12]. - The company is recognized as the top new media outlet in the AI and frontier technology sector according to third-party data platforms [12].

Core Viewpoint - The article discusses a remarkable case where AI was utilized to develop a personalized mRNA cancer vaccine for a dog named Rosie, diagnosed with a severe form of cancer, leading to significant improvement in her health [3][9][24]. Group 1: Background and Diagnosis - Rosie, a previously active dog, was diagnosed with a highly malignant and almost untreatable rare cancer after showing symptoms of lethargy and swelling [9]. - Traditional surgical options were deemed ineffective, and there were no suitable targeted drugs available on the market [10]. Group 2: AI Intervention - The dog's owner, Paul, a tech professional, decided to leverage AI to explore treatment options [4][11]. - ChatGPT provided insights into biological concepts and suggested immunotherapy, guiding Paul towards genetic sequencing [12][13]. Group 3: Development of mRNA Vaccine - After obtaining Rosie's genetic sequencing data, Paul used his expertise to analyze the data and identify potential targets for treatment [14][15]. - Despite initial setbacks in obtaining human-related immunotherapy drugs, Paul collaborated with UNSW RNA Research Institute to create a custom mRNA vaccine for Rosie [18]. Group 4: Treatment Outcomes - The mRNA vaccine was administered in two doses at the end of 2025 and early 2026, resulting in a 50% reduction in the tumor size within weeks [24][26]. - Rosie showed significant improvement in energy and health, even engaging in playful activities like chasing rabbits in the park [26]. Group 5: Ethical Considerations - The development and use of the mRNA vaccine underwent a rigorous ethical approval process, taking three months and requiring extensive documentation [31][32]. - Paul emphasized the importance of maintaining ethical standards in technology to ensure safe and beneficial outcomes [34].

Core Viewpoint - The article discusses the introduction of LabClaw, an AI-driven tool designed to automate various aspects of scientific research, significantly enhancing efficiency and reducing the workload for researchers [1][4]. Group 1: LabClaw Overview - LabClaw is described as a "skill package" for AI, enabling researchers to utilize over 200 skills related to biomedical research with a simple command [3][12]. - The skills are categorized into various fields, including Biology & Life Sciences, LabOS & Automation, Pharmacy & Drug Discovery, and more, with a total of 211 skills available [12]. - Each skill provides guidance on when and how to use it, as well as the expected outcomes, facilitating a streamlined research process [12]. Group 2: Functionality and Applications - LabClaw can automate entire workflows in research, such as single-cell analysis, drug discovery, and clinical trials, by calling upon relevant skills based on the research topic [14][15]. - It can function as an "Always-On Lab Agent," continuously monitoring experimental data and automatically triggering analysis and reporting when anomalies are detected [19][24]. - The integration of LabClaw with LabOS, a dedicated operating system, enhances its capabilities, allowing for real-time collaboration between AI and researchers [28][30]. Group 3: Development and Support - LabClaw is supported by prominent institutions, including Stanford and Princeton, and has received backing from NVIDIA [6][30]. - The development team includes notable figures in the fields of genetics and AI, such as Professor Le Cong and Professor Mengdi Wang, who have significant academic credentials and contributions to their respective fields [35][38]. Group 4: Impact on Research - The introduction of LabClaw is expected to lower the barriers to AI-assisted research, making it accessible with just a single command [44]. - The system's extensibility allows for the addition of new skills as research needs evolve, ensuring that it remains relevant and useful in a rapidly changing scientific landscape [43].

Venus团队 投稿 量子位 | 公众号 QbitAI 你随手拍下一张照片， AI 也许只会夸"真好看"，却说不出一句真正有用的建议。 面对构图失衡、主体模糊的照片，现有大模型往往停留在泛泛而谈的"赞美式反馈"上：既识别不了问题出在哪里，也无法给出具体、可操作的 拍摄指导。 针对这一挑战，北京大学彭宇新教授团队在美学理解领域开展了最新研究，定义了 美学指导 这一任务，并构建了首个美学指导数据集 AesGuide 。该数据集包含超过一万张照片，以及与之配套的专业分析和拍摄建议。在此基础上，团队进一步提出美学指导大模型 Venus ， 通过渐进式审美问答与思维链裁剪推理赋予大模型美学理解能力，使 AI 从"被动描述图像"迈向"主动指导拍摄"。 相关论文已被 CVPR 2026 接收，并已开源。 △ 图1. 美学指导任务示意图 从"图像描述"到"摄影指导" 智能手机的普及使拍照融入日常生活中，成为人们留存记忆、分享生活、记录情绪的便捷方式。但"拍得到"不等于"拍得好"，由于缺乏专业的 摄影经验与审美训练，许多用户在构图布局、取景视角与人景关系等关键环节难以做出准确判断，导致照片无法拍好，在质感与表现力上与专 业摄影 ...

LATENT团队 投稿 量子位 | 公众号 QbitAI 在所有运动场景中， 网球 几乎是人形机器人最难的一道考题： 高速来球逼迫瞬时判断，全身协同决定回球质量，满场奔跑则持续考验爆发力与控制力。 那么，当机器人真正站上球场，它能否像人类运动员一样完成判断、移动与连续回合击球？ 画面中，机器人迅速移动脚步调整站位，上下半身协同挥拍击球，并将球精准回击到指定位置。面对各种来球，它能够持续调整身体姿态与击 球时机，与不同水平的网球对手完成多回合连续对拉。 在网球这样的高动态、高对抗环境中，机器人面对的是时速超过几十公里的来球、变幻莫测的落点轨迹，以及对手不断变化的击球节奏。 更重要的是，这一能力并非依赖预编程动作实现，而是机器人通过深度强化学习自主习得—— 全球首次 在人形机器人上实现高动态网球对 打， 机器人正在实现从"机械复刻动作"向"智能决策响应"的底层跨越 。 研究团队提出了一种新的机器人运动学习方法，使人形机器人能够从不完美的人类动作数据中学习复杂的运动技能，并在真实世界中完成高动 态、高敏捷的网球击球与对打任务。 这背后，是来自银河通用与清华大学联合提出的新研究： LATENT （Learning A ...

Core Viewpoint - The discussion around AI is becoming increasingly polarized, with rapid technological advancements on one side and growing anxiety about AI's impact on jobs on the other [1][3]. Group 1: AI Technological Advancements - AI models are becoming more capable, with tools like OpenClaw being able to execute tasks and interact with systems [2]. - Bilibili has initiated a live challenge called "Lobster Replacing Humans," where AI is tested in real-world tasks through live streaming, allowing for an unedited view of its performance [5][6][11]. Group 2: Real-World Testing and Challenges - The live series involves various content creators assigning real tasks to OpenClaw, aiming to observe its performance in a pressure-testing environment [8][9]. - The first live test revealed mixed results, with initial tasks failing but later successes, such as developing an upgraded app, showcasing AI's potential and limitations [15][16][17]. Group 3: Future Live Challenges - Upcoming live sessions will explore AI's role in video production and commercial sales, with specific goals like selling 1 million units in a day [19][21]. - Additional challenges will test AI's capabilities in gaming and company management, further examining its practical applications [23][24]. Group 4: Broader Implications - The series serves as a public experiment to assess AI's real-world functionality, moving beyond emotional discussions to tangible results [30][34]. - The outcomes of these tests may provide clearer insights into AI's potential to replace human roles and its effectiveness in various industries [31][36].