Core Viewpoint - The article discusses the introduction of CLM (CPU-offloaded Large-scale 3DGS training), a system that allows for city-scale 3D reconstruction using a single consumer-grade GPU, specifically the RTX 4090, by offloading memory-intensive parameters to CPU memory, significantly lowering hardware requirements for large-scale neural rendering [1][21]. Group 1: 3D Gaussian Splatting (3DGS) Challenges - 3DGS has become a crucial technology in neural rendering due to its high-quality output and rendering speed, but it faces significant challenges when applied to complex scenes like urban blocks, primarily due to GPU memory limitations [2]. - A high-precision 3DGS model typically contains tens of millions to over a hundred million Gaussian points, with each point requiring substantial memory for parameters, gradients, and optimizer states. Even high-end GPUs like the RTX 4090, with 24GB of memory, can only handle about 15-20 million points, which is insufficient for city-scale scenes [2][3]. Group 2: CLM Design Principles - CLM is based on the observation that only a small fraction of Gaussian points are actively used during each rendering pass, with less than 1% of points accessed in large scenes [3]. - The system design of CLM involves dynamically loading Gaussian parameters from CPU memory as needed, rather than keeping all parameters in GPU memory [4]. Group 3: Key Mechanisms of CLM - **Attribute Segmentation**: CLM retains only "key attributes" (10 parameters) necessary for visibility checks in GPU memory, while the remaining 80% of "non-key attributes" are stored in CPU memory and loaded on demand [6][7]. - **Pre-rendering Visibility Culling**: Unlike traditional methods, CLM calculates visible Gaussian point indices before rendering, reducing unnecessary GPU computations and memory usage by only loading visible points from CPU memory [9][10]. - **Efficient CPU-GPU Collaboration**: CLM employs a multi-layered design to mitigate data transfer delays, including micro-batching, caching mechanisms, and intelligent scheduling to maximize efficiency and minimize communication overhead [12][13][14][15]. Group 4: Performance Results - CLM technology significantly increases model size, allowing for the training of 102.2 million Gaussian points on the "MatrixCity BigCity" dataset, a 6.7-fold increase compared to traditional methods, which maxed out at 15.3 million points [16]. - The quality of reconstruction improves with more parameters, achieving a PSNR of 25.15dB for the 102.2 million point model, compared to 23.93dB for the smaller model [18]. - Despite communication overhead, CLM maintains a training throughput of 55% to 90% of the enhanced baseline on the RTX 4090, and up to 86% to 97% on the slower RTX 2080 Ti [19]. Group 5: Broader Implications - CLM represents a significant advancement in addressing deployment bottlenecks in 3DGS training, integrating CPU resources into the training process without the need for multi-GPU setups, thus providing a cost-effective solution for large-scale scene reconstruction [21]. - The growing demand for efficient and low-cost 3D reconstruction tools in applications like digital twins and large-scale map reconstruction highlights the importance of CLM's approach in optimizing existing computational resources [21].

Core Viewpoint - The year 2025 is anticipated to be the "Agent Year," marking a significant shift in the industry towards practical applications of Agent technology [1][2]. Group 1: Development and Challenges of Agents - The Agent technology has transitioned from a nascent stage to practical engineering applications throughout the year [3][7]. - Key challenges in the implementation of Agents include the need for a robust engineering approach to manage complex systems and the importance of Agent Infrastructure (Infra) [6][21]. - The industry recognizes the value of Agents as they effectively address real-world problems, moving from theoretical discussions to tangible applications [6][12]. Group 2: Perspectives from Industry Leaders - Industry experts highlight a clear divide between traditional narratives from Silicon Valley and practical applications seen in smaller businesses, indicating a shift towards realism in Agent development [8][10]. - The emergence of AI coding tools is noted as a significant development, changing software engineering paradigms and serving as a universal interface for Agents [7][34]. - The consensus among experts is that the capital market is seeking new organizational methods, as the previous internet era's benefits have been largely exhausted [12][13]. Group 3: Engineering and Infrastructure - The concept of Agent Infra is crucial for managing the uncertainties inherent in Agent systems, with a focus on creating a safe and effective operational environment [21][22]. - The development of safety sandboxes and observability tools is essential for addressing the risks associated with autonomous Agent operations [22][23]. - The distinction between essential complexity and incidental complexity in enterprise problem-solving is emphasized, with a focus on building a common subset of solutions for various challenges [27][28]. Group 4: Future Trends and Directions - Future developments in Agent Infra are expected to focus on ensuring safe and reliable operations while optimizing the intelligence of Agents through continuous data utilization [38][39]. - The integration of memory management and semantic context is highlighted as a key area for enhancing Agent capabilities [40]. - The industry anticipates a significant transformation in mobile development ecosystems as Agents become mainstream, necessitating a shift in development methodologies and collaborative practices [41][44].

加入我们，你可以获得： 以下是岗位详情： 所有岗位不同能力层级职位均在开放，欢迎结合个人履历和经验申请。 编辑部 发自 凹非寺 量子位 | 公众号 QbitAI AI热潮还在汹涌，但如果你还不知道如何参与……那为什么不来 量子位 呢？ 我们是一家以 追踪AI新进展 为核心的内容平台，经过8年积累，目前拥有顶流影响力，广泛且备受认可的产业资源，以及时代风口的最佳观 测和学习生态位。 目前，我们有 三大方向 岗位招聘，希望你是 （或者能成为） 这三个方向的内容专家： 岗位均为全职，工作地点：北京中关村。 岗位面向： AI产业方向 岗位职责： AI产业方向 ：关注基建层创新，包含芯片、AI Infra、云计算； AI财经方向 ：关注AI领域创投和财报，跟踪产业链资本动向； AI产品方向 ：关注AI在应用和硬件终端方向的进展。 社招：覆盖编辑、主笔、主编各个层级，按能力匹配岗位； 校招：应届毕业生，接受实习且可转正。 站在AI浪潮之巅 ：第一时间接触和了解AI领域最新技术和产品，构建完整的AI认知体系。 玩转AI新工具 ：将各种AI新技术、新工具应用于工作，提升工作效率和创造力。 打造个人影响力 ：通过撰写独家原创内 ...

Core Viewpoint - The article discusses the challenges and opportunities in the medical AI sector, particularly focusing on the company Yunzhisheng, which has established itself as a leading player in the field by successfully integrating AI solutions into hospitals and demonstrating significant operational efficiency improvements [1][9][64]. Group 1: Medical AI Challenges - Patients increasingly consult AI chatbots before visiting doctors, leading to communication challenges in clinical settings [2][3]. - The high "hallucination rate" of general AI models in medical contexts can reach up to 40%, raising concerns about their reliability [4][5]. - Medical AI must navigate stringent requirements for stability, acceptance across various healthcare systems, and the high costs associated with medical errors [12][18][19]. Group 2: Yunzhisheng's Position - Approximately 40% of top-tier hospitals in China have adopted Yunzhisheng's medical AI solutions, indicating its strong market presence [9][22]. - The company has deployed its solutions in 400 hospitals, with a nearly 90% direct usage rate of generated medical records, significantly reducing doctors' time spent on documentation [22][23][25]. - Yunzhisheng's medical AI solutions are designed to integrate seamlessly into existing workflows, enhancing efficiency without adding to the workload of healthcare professionals [58][60]. Group 3: Technological Advancements - Yunzhisheng's latest model, "Shanhai·Zhimed 5.0," employs a dual-core system capable of processing structured information and multimodal inputs, enhancing its diagnostic capabilities [34][36]. - The model's architecture includes a three-layer data paradigm that improves its understanding of medical contexts and reduces hallucination rates to below 3% [42]. - The company has consistently ranked at the top of medical AI evaluation platforms, demonstrating its technological superiority [45][46]. Group 4: Business Growth and Market Trends - Yunzhisheng's medical business revenue reached 0.70 billion, a 22.3% increase year-on-year, highlighting its growth trajectory [66]. - The average revenue per medical client has more than doubled, indicating a significant increase in customer value [66]. - The broader market for medical AI is expected to grow, with increasing investments and policy support aimed at integrating AI into healthcare workflows [79][82][86].

Core Viewpoint - The article discusses the launch of Huawei's nova 15 series, highlighting its innovative features, design, and pricing strategy, which aims to attract a younger audience while enhancing AI capabilities in photography and communication. Product Overview - The nova 15 series includes three models: standard, Pro, and Ultra, all equipped with HarmonyOS 6 [4] - The Ultra and Pro versions feature a horizontal stacked design and dual star lens module, resembling "two big eyes" [5] - The series introduces the Kirin 9 series chip for the first time in the Pro and Ultra versions, aligning performance with Mate and Pura series [6] - Pricing starts at 4199 yuan for the Ultra version, 3499 yuan for the Pro version, and 2699 yuan for the standard version [7][10] AI Capabilities - The nova 15 series incorporates advanced AI features, particularly in photography, with the Ultra and Pro versions featuring a dual red maple imaging system [14] - Color accuracy has improved by 120%, and spatial resolution has increased by 100,000 times due to the red maple original color lens [17] - AI assists in photo composition and includes a unique "color dipping" feature that allows users to apply colors and styles from online images to their own photos [21][22] - The series also offers AI-driven photo editing capabilities and one-click content creation [25][28] Communication Features - The phone includes a call summary feature that automatically generates key points after a call, which can be synced to a memo [31] - Dual-direction call noise reduction is optimized for noisy environments like subways and malls [33] - A family fraud prevention feature allows family members to share risk information and assist in handling suspicious calls [35] Design and Specifications - The nova 15 series maintains a recognizable design with a focus on youthfulness and high identification [39] - The Ultra version features a 2.5D flat screen and is available in four colors: "vibrant green," "good match purple," "zero degree white," and "fantasy night black" [40] - Both the Ultra and Pro versions come with a 6500mAh battery and support for 50W wireless fast charging [42][51] - The Ultra version has a thickness of 6.8mm and weighs approximately 209g [43] - The rear camera system includes three 50MP RYYB lenses, supporting variable aperture and optical stabilization [45] - The series is equipped with Kunlun glass and supports IP68 & IP69 dust and water resistance [47] Performance Improvement - According to Huawei's lab tests, the overall performance of the nova 15 series has improved by 62% compared to the previous generation [53]

AIME 25和人类最后 考试 （HLE） 等基准 中，GLM-4.7分数超GPT-5.1； SWE-Bench分数达（73.8%，+5.8%），创开源新高。 鱼羊 henry 发自 麦蒿寺 量子位 | 公众号 QbitAI 2025倒计时，新SOTA模型涌现没有放缓迹象。 一夜之间，编程SOTA模型易主，而且上线即开源，依然来自中国大模型公司—— 智谱AI，GLM-4.7。 这波更新，技术报告里满眼都是 Coding ， Coding ，还是 Coding 。 而能力的提升，带来的最直观效果是： 官方Demo显示，写个植物大战僵尸不费劲： 官网Chatbot和API均已就为，现在就能在线开玩。 Demo来吧，展示 在前端生成质量上，GLM-4.7展现出明显升级：页面结构更干净、组件层级更清晰。 相比GLM-4.6，更像是现代的Web UI，网友元素中更加美观。 总而言之，模型这么一发，双旦的节庆氛围一下到位了（doge）。 在复杂几何结构与空间关系的表达上，GLM-4.7模型能够保持较好的结构一致性与细节稳定性。 3D资产的生成质量也有显著提升。 在PPT与视觉物料生成方面，GLM-4.7标题层级明确、元素 ...

Core Viewpoint - The article discusses the limitations of AI agents in real-world applications compared to their impressive demonstrations, emphasizing that adaptability is a key factor for improvement [1]. Summary by Sections Definition and Functionality of Agents - Agents are defined as AI systems that can plan, utilize tools (such as search engines and databases), and remember information to complete complex tasks independently [3]. Adaptability Framework - The core bottleneck in current agent systems is adaptability, specifically how models adjust their behavior based on feedback signals [6]. - A 2x2 classification framework is proposed to categorize existing adaptation methods into four paradigms based on two dimensions: who is optimized (the agent or the tools) and where the feedback signal comes from (tool execution results or agent output evaluations) [7][8][9]. Four Paradigms of Adaptation - **A1 Paradigm**: Agents learn from feedback based on tool execution, such as whether code runs successfully [10]. - **A2 Paradigm**: Uses the agent's final output as the optimization signal, exemplified by models like DeepSeek-R1 that train reasoning capabilities through reinforcement learning [11]. - **T1 Paradigm**: Tools are pre-trained independently and then called by the agent, allowing for plug-and-play functionality [12]. - **T2 Paradigm**: Tools optimize themselves based on the agent's output, creating a symbiotic relationship [13]. Benefits of Classification - This classification helps developers avoid trial and error when improving AI capabilities, allowing for targeted adaptations based on specific needs [15]. - It also clarifies trade-offs: modifying AI (A1/A2) is flexible but costly, while modifying tools (T1/T2) is cheaper but limited by the AI's inherent capabilities [16]. Key Findings on Data Efficiency - The T2 paradigm demonstrates significantly higher data efficiency compared to the A2 paradigm. For instance, the Search-R1 using A2 requires approximately 170,000 training samples, while T2 only needs 2,400 samples, achieving comparable results [18][19][20]. Frontiers in Adaptability Research - The article identifies four cutting-edge directions for agent adaptability research: - **Co-Adaptation**: Aims for agents and tools to optimize together within the same learning cycle, presenting challenges in credit assignment [21]. - **Continual Adaptation**: Addresses the need for agents to continuously learn new skills without forgetting old ones in a changing environment [23]. - **Safe Adaptation**: Highlights concerns that large models may erode safety measures established during supervised fine-tuning, making them more vulnerable to attacks [25]. - **Efficient Adaptation**: Focuses on resource-constrained scenarios, discussing techniques like LoRA and FlashRL for efficient learning [27]. Additional Resources - The article mentions that a GitHub repository has been opened to continuously collect related papers and resources, serving as a guide for developers building agent systems [29].

一凡 发自 凹非寺 量子位 | 公众号 QbitAI 一次大规模停电，暴露了全球无人车一哥的短板。 被曝估值冲上千亿美元没几天，Waymo就因为当地停电全面停摆了，挡在路中间，造成城市 拥堵，相关视频疯传。 马斯克第一时间"补刀"，表示自家Robotaxi就没受到影响。看上去，特斯拉代表的L2渐进式 路线，似乎小胜了一局……反正马哥认为这就是彰显优越性的时刻。 在Robotaxi战场上，今年马斯克的一举一动，都把自动驾驶之争推向了新的高潮，大洋两岸 更多玩家开始入场，沿着「特斯拉路线」前进，与「Waymo路线」争夺自动驾驶圣杯。 所以问题是，停电是如何影响Waymo Robotaxi的？ 当地停电，Waymo停工 Waymo停摆源自一场火灾，旧金山变电站失火，导致当地大规模停电，据说直接影响到13 万居民用电。 更要命的是，因为大范围停电，马路上的红绿灯都不亮了，引发Waymo无人车全面停摆。 真是屋漏偏逢连夜雨，本就混乱的交通，这下因为无人车挡在路上变得更堵了。Waymo只好 连夜找拖车运走了无人车，同时宣布在当地停运，目前还不清楚什么时候重新上线。 所以为啥停电会导致Waymo停运？首先是官方回应暴露的运 ...

编辑部 整理自 MEET2026 量子位 | 公众号 QbitAI 从大模型智能的"语言世界"迈向具身智能的"物理世界"，仿真正在成为连接落地的底层基础设施。 在本次量子位MEET2026智能未来大会上，光轮智能联合创始人兼总裁 杨海波 给出了他的观察： 具身智能的规模远大于文本与视觉模型，因为数据维度更真实、更复杂。 这也就意味着，具身智能时代的核心，不是算法本身，而是它所依赖的数据是否有效、可扩展——仿真是唯一能够解决数据问题的方案。 在仿真策略的路上，会遇到仿真不真实、Sim2Real不可靠等行业痛点， 光轮智能正在通过自研的一整套"测量、生成、求解"仿真基础设施来 解决这些问题 ，为具身智能提供数据、训练、评测的全流程解决方案。 △ 杨海波指出光轮智能深耕合成数据领域 另外杨海波还进一步指出， 仿真不是孤立的技术工具，需要以真实产业需求为锚点，通过应用场景构建生态。 其中， 具身仿真资产制作是生态的源头活水 ，依托自动化物理测量与生成技术，产出高物理真实的规范化数据资产，为具身训练提供核心燃 料； 大规模RL训练则通过并行的虚拟场景让智能体高效试错学习，将数据价值转化为具身实际技能 ，同时反向打磨仿真 ...

Core Insights - Gemini 3 Flash outperforms its predecessor Gemini 2.5 Pro and even the flagship Gemini 3 Pro in various benchmarks, achieving a score of 78% in the SWE-Bench Verified test, surpassing Gemini 3 Pro's score of 76.2% [1][6][9] - The performance of Gemini 3 Flash in the AIME 2025 mathematics competition benchmark is notable, scoring 99.7% with code execution capabilities, indicating its advanced mathematical reasoning skills [7][8] - The article emphasizes a shift in perception regarding flagship models, suggesting that smaller, optimized models like Flash can outperform larger models, challenging the traditional belief that larger models are inherently better [19][20] Benchmark Performance - In the Humanity's Last Exam, Flash scored 33.7% without tools, closely trailing Pro's 37.5% [7][8] - Flash's performance in various benchmarks includes: - 90.4% in GPQA Diamond for scientific knowledge [8] - 95.2% in AIME 2025 for mathematics without tools [8] - 81.2% in MMMU-Pro for multimodal understanding [8] - Flash's speed is three times that of Gemini 2.5 Pro, with a 30% reduction in token consumption, making it cost-effective at $0.50 per million tokens for input and $3.00 for output [9] Strategic Insights - Google’s team indicates that the Pro model's role is to "distill" the capabilities of Flash, focusing on optimizing performance and cost [10][12][13] - The evolution of scaling laws is discussed, with a shift from merely increasing parameters to enhancing reasoning capabilities through advanced training techniques [15][16] - The article highlights the importance of post-training as a significant area for future development, suggesting that there is still substantial room for improvement in open-ended tasks [17][18] Paradigm Shift - The emergence of Flash has sparked discussions about the validity of the "parameter supremacy" theory, as it demonstrates that smaller, more efficient models can achieve superior performance [19][21] - The integration of advanced reinforcement learning techniques in Flash is cited as a key factor in its success, proving that increasing model size is not the only path to enhancing capabilities [20][22] - The article concludes with a call to reconsider the blind admiration for flagship models, advocating for a more nuanced understanding of model performance [23]