育材班走到第七年，我们深度链接着近两百位新材料创业者，参与和感受着新材料创业的变化。 目前，七期学员的招募工作已经完成，共录取32名学员。由于大部分学员通过推荐进行报名，与育材 班匹配度较高，因此也是我们最难选择的一年，在此，感谢所有推荐人和报名者对育材班的信任！ 通过与每位报名者的深度访谈，我们感受到创业者有一些新变化，特别是与第一期育材班相比。 感受最深的是，学员的创业前经历越来越多样了，前几期育材班科研人员创业占比较高，本期报名的 学员，接近一半有大企业工作经验，对于商业、管理有很好的经验，对行业有很深的认知，也积累了 很多行业资源。科研人员创业的团队完整性也比较高，除了研发人员以外，也有做过企业运营、投资 等的合伙人共同创业。各种不同经历的人参与到新材料创业中来，背后既有商业机会的考量，也有个 人抱负的追求。 同时，我们也明显感受到了产业资源与创业项目的合力发展。有三分之一左右的创业项目通过各 种形式与产业资源进行了深度合作，有不少项目拿到了大企业的投资，成为大企业创新发展的突破力 量；科研人员创业也大多会优先寻找有产业背景的资金。大企业对于所投项目的赋能也有所体现，比 如母公司的专业管理团队为项目梳 ...

广东省科技厅二级巡视员黄江康在致辞中强调，联盟的成立是落实国家《新一代人工智能发展规划》的 重要举措，更是推动粤港澳大湾区建设"具有全球影响力的国际科技创新中心"的关键一步。 广州市委常委、南沙区委书记刘炜表示，南沙将聚焦平台建设、开放共享、场景应用与生态共赢，始终 以"创业者"的姿态、"合伙人"的诚意、"可预期"的环境，与联盟企业同心并肩、向新而行，带动更多产 业资源、优质项目落地南沙。 会上，多项人工智能项目集中签约。华为与香港科技大学(广州)启动"科教创新孵化中心"合作项目，产 教融合与科教融汇双向驱动；同时，华为与中铁隧道局合作规划"隧智汇"隧道工程行业大模型体系架 构、搭建安全可控的大数据平台底座，同步深化数字化人才共育，推动隧道工程全流程数智化转型。 5月26日，由广东省科学技术厅、广州市南沙区政府指导，香港科技大学(广州)、广州市南沙区科学技 术局联合主办的"湾区人工智能产业创新联盟成立大会"在广州南沙举行。活动汇聚政府、高校企业及国 际专家代表400余人，通过战略签约、政策发布、智库共建、专家座谈等系列活动，推动粤港澳大湾区 人工智能产业协同创新。 大会现场，由香港科技大学(广州)与华为联合港 ...

Core Insights - Galaxy General, founded in May 2023, has rapidly emerged in the humanoid robot sector, showcasing its Galbot G1 robot with a task success rate of 99.97% during the 2025 Zhongguancun Forum [1][9] - The founder, Wang He, emphasizes the importance of synthetic data for training humanoid robots, arguing that relying solely on real data is costly and time-consuming [4][16] - The company aims to drive the practical application of embodied intelligent robots across various scenarios, focusing on real-world utility [10][15] Data Strategy - Galaxy General has established a self-developed synthetic data generation pipeline, allowing for the mass production of diverse synthetic data at low marginal costs, which constitutes over 99% of the training data [5][6] - The company employs a strategy of "synthetic pre-training + real data alignment," which is deemed cost-effective and sustainable even as humanoid robot shipments increase [5][16] - The GraspVLA model, developed in collaboration with academic institutions, showcases the effectiveness of synthetic data, achieving a record training volume of one billion frames of "visual-language-action" data [6][7] Application and Development - The Galbot G1 robot is designed for versatility, capable of performing various tasks in multiple environments, including retail and healthcare [11][12] - The company has initiated strategic partnerships, such as with Suzhou, to implement humanoid robots in commercial retail and advanced manufacturing [10][14] - The "smart pharmacy" solution, which utilizes Galbot G1 for inventory management, is set to be deployed in approximately 100 stores in major cities by 2025 [13] Future Outlook - Galaxy General believes that the development of humanoid robots will continue to face challenges related to data acquisition and hardware capabilities, with a focus on achieving practical productivity [16][17] - The company anticipates significant breakthroughs in humanoid robot capabilities by 2025-2026, potentially leading to widespread adoption in various sectors [17]

论文第一作者林宏彬来自香港中文大学(深圳)理工学院的Deep Bit 实验室、深圳市未来智联网络研究院，导师为李镇老师。目前实验室的研究方向包括：自动驾 驶、医学成像和分子理解的多模态数据分析和生成等。 论文标题： DriveGEN: Generalized and Robust 3D Detection in Driving via Controllable Text-to-Image Diffusion Generation 论文链接： https://www.arxiv.org/abs/2503.11122 GitHub： https://github.com/Hongbin98/DriveGEN 任务背景 随着新能源汽车产业的持续发展，智能驾驶辅助技术的应用越来越广泛。其中，基于纯视觉的自动驾驶方案只需使用多视角图像进行环境感知与分析，具有 成本低、效率高的优势，因而备受关注。然而在实际应用中，视觉感知模型的泛化能力至关重要。 来自香港中文大学（深圳）等单位的学者们提出了一种名为 DriveGEN 的无训练自动驾驶图像可控生成方法。该方法无需额外训练生成模型，即可实现训 练图像数据的可控扩充，从而以较 ...

Investment Rating - The report adjusts the investment rating of the company to "Accumulate-A" [4][10]. Core Views - The company reported a slight increase in revenue and profit for 2024, achieving revenue of 840 million yuan, a year-on-year growth of 2.8%. The transmission product line saw a revenue decline of 3.5% to 720 million yuan, primarily due to a slowdown in domestic transmission network investments. In contrast, the data and access product line revenue surged by 70% to 100 million yuan, driven by strong demand for DCI subsystem products since the second half of 2024 [4][5]. - The first quarter of 2025 showed a slight revenue increase to 200 million yuan, but net profit decreased by 36.8% year-on-year to 20 million yuan, reflecting challenges in performance expectations [4][5]. - The DCI market in North America is experiencing strong demand, with the company holding competitive advantages in DCI optical and electrical layers. The company has completed the development of a full range of DCI equipment and is actively expanding production capacity [6][10]. Financial Data Summary - The company’s projected revenue for 2025 is 1.25 billion yuan, with net profit expected to reach 139 million yuan, reflecting a year-on-year growth of 38% [10][12]. - The gross margin is expected to improve gradually, with projections of 31.7% in 2025 and 33.3% in 2026 [12][14]. - The company’s earnings per share (EPS) is forecasted to be 1.15 yuan in 2025, increasing to 2.38 yuan by 2027 [12][14].

Group 1 - The core viewpoint of the article is that Guangdong has introduced policies to promote the high-quality development of the online gaming industry, focusing on technological advancements and collaboration with educational institutions [1] Group 2 - The policies encourage enterprises to conduct research and development around "bottleneck" technologies in the gaming sector, particularly in areas such as virtual engine development and data security [1] - Support is provided for the application of cutting-edge technologies like artificial intelligence in game development and the transformation of advanced gaming technologies to other fields [1] - The provincial government will allocate funds to support high-quality projects that achieve significant breakthroughs in gaming technology, with a maximum one-time subsidy of 5 million yuan available [1]

Core Viewpoint - The article emphasizes the importance of balanced diet and exercise for achieving a healthy weight, while addressing common misconceptions about exercise that could hinder weight loss efforts and pose health risks [1] Group 1: High-Intensity Interval Training (HIIT) - HIIT is effective in improving cardiovascular function, body composition, and metabolic risk factors in obese individuals, but lacks evidence proving it is more effective than traditional aerobic exercise [2] - Health screening is recommended for overweight individuals before engaging in HIIT [2] Group 2: Fasting Exercise - Exercising on an empty stomach is more harmful than beneficial for overweight individuals, as it can lead to muscle breakdown and various health issues due to low blood sugar levels [3] Group 3: Sweating Suits - The use of sweating suits for exercise is dangerous, primarily resulting in water loss, which can lead to dehydration and heat-related illnesses, especially in overweight individuals [4] Group 4: Weekend Warriors - The practice of intense exercise only on weekends, known as "weekend warriors," is discouraged due to significantly increased risks of acute myocardial infarction and musculoskeletal injuries compared to regular exercisers [5] Group 5: Recommended Exercise Practices - Regular, gradual exercise aligned with the body's biological clock is advised to enhance effectiveness, with suggestions to train during times of higher cortisol levels for improved performance [6] - Aerobic exercises such as walking, swimming, and using stationary bikes are recommended for overweight individuals, while those with joint issues should consider low-impact activities like swimming [6]

机器人数据仿真专家 20250521 摘要 VLA 仿真在机器人感知深度学习中应用广泛，但在真实场景迁移效果差， 主要受限于图像真实感和物理参数模拟的挑战，现阶段更适用于算法原 型验证。 机器人训练中，传感器仿真、物理交互和场景重建是常用的数据生成方 法，但高逼真图像生成和精确物理参数模拟仍是难题，限制了模型在真 实世界的泛化能力。 仿真器在机器人训练中优势明显，尤其在电信号仿真方面，但感知层面 的数据分布差异导致环境交互效果不佳，仿真可迁移性取决于任务的数 据分布差距。 通过观看视频训练自动驾驶系统和机器人面临模态差异和重建精度问题， 难以实现完整动作或任务流的学习，视觉语言规划可作为辅助手段。 当前主流数据采集及训练方式依赖真实量产数据闭环，机器人领域则依 赖仿真器，但仿真数据训练的模型在真实世界中泛化性较差。 数据而非模型是当前主要挑战，硬件不统一和数量不足导致数据量少且 质量参差，跨本体数据运用受限，需标准化硬件尺寸和旋转比例以提高 数据利用效率。 人形机器人在工业场景中的价值在于通用性，而非精细化操作，通过解 耦方式采集数据，标准化传感器并解耦本体，可确保数据的共用性。 Q&A 仿真数据在机器人任务 ...

|2025年5月20日 星期二| 每经记者｜赵雯琪 每经实习编辑｜余婷婷 NO.1 零一万物回应联合创始人谷雪梅已离职 5月19日，有市场消息称，零一万物联合创始人谷雪梅在近期离职，正在筹备创业。在零一万物期间， 她主要负责模型预训练和C端产品。对此，零一万物向《每日经济新闻》记者确认，谷雪梅作为预训练 负责人已在数月前离职。据公开消息，谷雪梅是谷歌中国的第一位本土女工程师，主要关注大规模分布 式计算平台、搜索推荐及广告、知识图谱等方面的技术研发。 点评：当下直播行业竞争白热化，增长放缓进入存量博弈。头部MCN（多频道网络）纷纷收缩多元战 线、加大技术投入转型。此次三只羊旗下多家公司法定代表人变更，或为内部管理优化与战略调整之 举。新负责人可能带来新理念，更加聚焦核心业务，但也同样面临战略定力和战术上的挑战。 NO.3 宠物综合O2O服务平台宠贝贝旗下品牌获战略投资 5月19日，宠物综合O2O（线上到线下）服务平台宠贝贝宣布，旗下子品牌猫王谷天使轮获得天使投资 人成晓华的战略投资。猫王谷定位为爱宠全生命周期服务平台，已在佛山开设了一家线下门店，第二家 门店将于6月花落深圳南山区。猫王谷预计年内直营门店将超10 ...

Core Insights - The co-founder of Zero One Everything, Gu Xuemai, has left the company several months ago to prepare for a new venture [1] - Zero One Everything has shifted its business strategy since Q4 of last year, focusing on lightweight model training, fine-tuning, and practical applications of AI [1] - The company’s COO, Huang Huiwen, stated that they will no longer develop ultra-large parameter base models, but will instead concentrate on AI market applications [1] - As of the beginning of 2025, Zero One Everything's total revenue has already surpassed the entire revenue of the previous year [1]