（原标题：教机器人"做人"：广东举办了一场机器人的"教练大赛"） 南方财经记者郑康喜 实习生 钟宸 广州报道 12月13日，在2025广东省人工智能与机器人技能大赛现场，几十具机器人不仅爬坡过坎、跌跌撞撞，也 在模拟工业场景中的精准作业与系统集成包括设备巡检、零部件装配等任务。 赛中，有机器人因失去重心向前栽倒，脑袋滚落一旁；有机器人稍有不慎便会跑偏。但比赛也不乏强悍 的机器人参赛。来自联通（广东）产业互联网有限公司的选手凭借良好的产品性能和技能能力，拿到机 器人工业特种应用赛道第一名。 一场技能竞赛，机器人的机械结构设计、运动控制算法等各项性能被全方位拷问，其在制造、交通、教 育等核心场景的融合应用程度也被立体化检验。更为重要的是，让这些钢铁之躯拥有"灵魂"的训练师， 不断探索人机交互的边界，推动着技术的革新和产业化进程。 竞赛的背后，是广东对当下机器人产业核心痛点的精准思考：当前，机器人尤其是人形机器人的能力， 还不足以满足产业化应用的需求，具身智能的进阶速度远落后于大语言模型。背后的原因在于缺乏大量 场景训练。但目前，行业内诸多产品缺乏后半段真实场景的训练，其数据仍依赖通用大模型，这就使得 机器人无法完 ...

Industry Overview - The humanoid robot industry is accelerating mass production by 2025, with companies like Zhiyuan achieving deliveries in the thousands and a significant increase in the localization rate of core components [1] - Continuous support from domestic and international policies is driving both market enthusiasm and the commercialization process [1] - The industry trend is expected to provide dual drivers for performance and valuation in the sector, according to a report from CITIC Securities [1] Market Performance - Siling Co., Ltd. has seen a maximum stock price increase of 329% year-to-date, while Haoshi Electromechanical has experienced a 134% increase [1] Component Cost Analysis - In the humanoid robot supply chain, the cost breakdown for various components includes: - Actuators: 1% cost share, with companies like Sanhua Intelligent Control and Top Group involved [2][3] - Frameless torque motors: 4% cost share, with companies such as Buke Co., Haoshi Electromechanical, and Leisai Intelligent [2][3] - Harmonic reducers: 5% cost share, with companies like Green Harmonic, Hanyu Group, Siling Co., and Fengli Intelligent [2][3] - Encoders: 2.7% cost share, with companies including Inovance Technology and Aopu Optoelectronics [2][3] - Bearings: 1.2% cost share, with companies like Lixing Co., Southern Precision, and Guoji Precision [2][3] Company Highlights - Haoshi Electromechanical is focusing on building a second growth curve through its robotics business, which includes a robust "N+1+3" business system [4] - The company has achieved significant improvements in transmission precision and reduced noise levels in its products, positioning itself competitively in the market [4] - Buke Co. is recognized as a domestic leader in frameless torque motors, with projected sales of approximately 24,000 units in 2024 [5] - Siling Co. is a leading domestic company in harmonic reducers, with a focus on applications in industrial and humanoid robots [5] - Green Harmonic holds a 26% market share in the domestic harmonic reducer market, with robot-related revenue constituting over 80% of its total income [6]

Group 1 - Yushu Technology has launched its humanoid robot "App Store," allowing users to upload and share trained models for various actions such as dance and martial arts [1] - The platform includes data collected from real robots, enabling users to upload their own datasets or download others' for algorithm training and optimization, thus saving development time for both amateurs and professional teams [3] - The public beta version features preset actions including humorous movements, a dance, and Bruce Lee's martial arts, showcasing the company's advanced motion control capabilities [5] Group 2 - Yushu Technology, founded on August 26, 2016, focuses on the R&D, production, and sales of high-performance humanoid robots and dexterous robotic arms [6] - The company's annual revenue has exceeded 1 billion yuan, meeting the basic requirements for an IPO on the A-share market [7] - The IPO process is accelerating, with the China Securities Regulatory Commission indicating that Yushu Technology has completed its counseling phase and is in the "counseling acceptance" status, preparing to submit its IPO prospectus between October and December [8]

由立德机器人平台（机器人大讲堂）发起的「第六届中国机器人行业年会」，将于 2025年12月18日-19日在 浙江杭州滨奥皇冠假日酒店举办！ 大会以 "见证中国机器人领军企业力量，点亮具身智能时代变革新星"为主 题，聚焦人形机器人技术与具身智能价值，搭建政产学研投高端对话平台。依托杭州科创枢纽优势，助力产业 集聚与全球竞争主动权，共筑共赢生态！ 本届大会参展企业共计 21家， 第三批揭秘的6家展商分别为绿的谐波、晶品特装、青瞳视觉、PNP机器人、 安培龙、售后宝，六家企业 将携核心技术与前沿产品亮相年会现场。我们也将陆续带来展商展品速递资讯， 欢迎持续关注！ ▍ 参展企业展品介绍 （以下排名不分先后） 【 苏州绿的谐波传动科技股份有限公司 】 绿的谐波是全球领先的核心零件制造商，自 2003年起专注精密谐波传动技术创新研发，率先打破国外技术垄 断，在谐波减速器领域国内市占率第一、全球第二，入选全球人形机器人百强。2020年在科创板上市，作为 国家专精特新"小巨人"与制造业单项冠军，公司拥有160项专利，主导6项国家标准，产品达国际领先水平， 广泛应用于机器人、具身智能、智能制造等领域。 1. 谐波减速器 2. ...

天眼查App显示，近日，广东鼎泰智能技术有限公司成立，注册资本2000万人民币，经营范围为一般项 目：智能机器人的研发；机械设备研发；电机及其控制系统研发；新材料技术研发；五金产品研发；智 能机器人销售；工业机器人制造；工业机器人销售；工业机器人安装、维修；电子元器件与机电组件设 备制造；电子元器件与机电组件设备销售；电子专用设备制造；电子专用设备销售；特种陶瓷制品制 造；特种陶瓷制品销售；金属工具制造；金属工具销售；非金属矿物制品制造；非金属矿及制品销售； 真空镀膜加工；软件开发；软件销售；人工智能应用软件开发；货物进出口；技术进出口。（除依法须 经批准的项目外，凭营业执照依法自主开展经营活动）。 ...

Group 1: Investment and Financing Activities - Shanghai Zhishi Robotics completed A+ round financing of several tens of millions, exclusively invested by Yinfeng Capital, focusing on four-way shuttle robots for automated storage solutions [2] - Wuhan Jizhao Energy Technology announced the completion of several tens of millions in angel+ round financing led by Guanghe Venture Capital, aimed at supporting the delivery of the world's first MW-level liquid metal battery energy storage system [2] - Deep Intelligent Pharma announced nearly $50 million in D round financing led by Dinghui Baifu, with participation from existing shareholders, to enhance AI-driven drug development [4] - Deshengji Pharmaceutical completed $108 million in B round financing, with investors including IDG Capital and Sequoia China, focusing on precision therapeutic drug development [5] - Skild AI is in negotiations for over $1 billion investment from SoftBank and NVIDIA, potentially increasing its valuation to approximately $14 billion [6] - Pruy Benchmark completed over $10 million in new financing, focusing on AI-driven drug research and development [7] Group 2: Company Developments and Innovations - Zhishi Robotics aims to enhance storage density and efficiency in warehouses through innovative four-way shuttle technology [2] - Jizhao Energy is the first domestic supplier of liquid metal battery energy storage systems, with a pilot production line established for various battery products [2] - Shenzhen Zhaqi Xinzhi focuses on creating a high-throughput, low-cost protein detection platform using semiconductor technology [4] - Skild AI, founded by former Meta AI researchers, aims to develop general AI software systems for robotics, enhancing their perception and decision-making capabilities [6] - Pruy Benchmark leverages multi-omics and AI algorithms to support drug mechanism research and biomarker development [7]

12月12日至14日，"2025年粤港澳大湾区人工智能与机器人产业大会暨广东省人工智能与机器人技能大 赛"在广州举办。据悉，本次大会上，广东省具身智能训练场"1+1+N"体系将正式发布。"1+1+N"体系包 括："1"个广东核心中枢——广东省具身智能训练场（简称"省训练场"），由主训练场和管理中心两部 分构成。"1"个深圳示范窗口——深圳具身智能示范区，探索建设全国首个具身智能机器人友好示范城 区，侧重封闭训练后的机器人在街区开放式训练，激活场景价值。"N"个垂直分训练场——横跨不同地 市、深耕不同领域的垂直场景专业化训练平台，联动地市、赋能产业、精准定位、解决痛点，与主训练 场优势互补、资源共享、共建共创、相互赋能。（深圳发布） ...

客观上看，资本的密集涌入、市场的高度关注，对于鼓励创新来说是件好事，但真正的风险在于， 如果资本过度集中在少数"讲故事"的企业，过于强调投资市场回报而忽略了基础研发、核心零部件突 破、应用场景深耕等长期投入，那么当潮水退去，留下的可能并不是我们想要的坚实的产业基础。 原标题：理性看待机器人产业速度与泡沫 资本的密集涌入、市场的高度关注，对于鼓励创新来说是件好事。但真正的风险在于，如果过于强 调投资市场回报而忽略了基础研发、核心零部件突破、应用场景深耕等长期投入，那么当潮水退去，留 下的可能并不是我们想要的坚实的产业基础。 近来，有关人形机器人的声音不绝于耳，大概观点有两个：一是人形机器人市场将持续升温，明年 将成为人形机器人落地的关键一年；二是有市场人士担忧，当前人形机器人技术成熟度、应用场景、安 全性等方面仍有不少缺陷，市场过热或引发投资泡沫。 在机器人产业融资活跃、资本热捧之际，及时提醒市场保持清醒头脑与理性态度是十分必要的。正 如国家发展改革委新闻发言人在回应上述关切时所言，"速度"与"泡沫"一直是前沿产业发展过程中需要 把握和平衡的问题，对于具身智能产业也一样。 近年来，在创新引领和需求释放的双重作用 ...

Core Insights - The article discusses the ViVLA framework, which enables robots to learn new skills from single video demonstrations, addressing the limitations of existing Vision-Language-Action (VLA) models in generalizing to tasks outside their training distribution [1][2][25] Group 1: Challenges in Robot Skill Generalization - Four core challenges hinder the generalization of robot skills: insufficient fine-grained action recognition, differences in action representation and modalities, inherent flaws in autoregressive modeling, and a lack of diverse expert-agent pairing data [4][5][7] Group 2: ViVLA's Technical Framework - ViVLA employs a three-layer technical system: unified action space construction, parallel decoding optimization, and large-scale data generation to achieve efficient learning from single video demonstrations [8] - The first layer focuses on latent action learning through an Action-Centric Cycle-Consistency (A3C) framework to bridge the gap between different expert and agent action spaces [10] - The second layer enhances model training efficiency with parallel decoding and spatiotemporal masking strategies, improving video understanding and reducing inference delays [11][12] Group 3: Data Generation and Validation - ViVLA's data generation pipeline converts human videos into high-quality paired data, resulting in a dataset of over 892,911 expert-agent training samples [13][17] - The framework's effectiveness is validated through a three-tier performance verification system, demonstrating significant improvements in unseen task success rates compared to baseline models [14][16] Group 4: Performance Metrics - In the LIBERO benchmark, ViVLA achieved over a 30% performance increase in unseen tasks compared to baseline models, with success rates of 74% in real-world manipulation tasks, significantly outperforming other models [14][16][18] - The model maintained a success rate of over 70% in varying environmental conditions, showcasing its robustness [20] Group 5: Future Directions and Limitations - While ViVLA represents a breakthrough in single-sample video imitation learning, there are areas for optimization, including enhancing error recovery capabilities and expanding data diversity through automated filtering of human videos [25][27]

Core Insights - Guangdong's artificial intelligence and robotics industry is rapidly advancing, highlighted by the recent showcase of innovative technologies at the 2025 Guangdong-Hong Kong-Macao Greater Bay Area AI and Robotics Industry Conference [1][4] - The establishment of the Guangdong AI and Robotics Industry Alliance aims to foster collaboration among enterprises, universities, and research institutions, focusing on ten key initiatives to enhance industry development [2][3] Industry Development - The Guangdong AI and Robotics Industry Alliance was formed in June, bringing together various stakeholders to promote collaboration and innovation within the industry [2] - The alliance's "Ten Ones" initiative includes the development of an industry index, a comprehensive industry chain map, a white paper, application scenarios, an innovation and entrepreneurship competition, an award, an industry conference, an expo, an information platform, and an industry fund [2][3] Competitive Advantage - Guangdong's unique "soft-hard integration" capability is a significant advantage, providing vast industrial data for AI and a robust hardware supply chain for robotics [3] - The province leads the nation in the production of smart terminals, with 40.9% of mobile phones, 24.6% of computers, 57.7% of televisions, and 35.3% of smartwatches produced in 2024 [3] - Guangdong accounts for 44% of China's industrial robot production, maintaining its position as the leading manufacturing base and supply chain hub for the robotics industry [3] Technological Breakthroughs - Guangdong has made significant advancements in humanoid robots and industry-specific AI models, with notable achievements in generative AI and core technologies like AR waveguides and high-precision force control [4][5] - The province has completed 105 registrations for generative AI services, representing approximately 17% of the national total, showcasing its leadership in the sector [5] Commercialization and Application - Leading companies in Guangdong, such as GAC and Xpeng, have successfully commercialized advanced technologies in smart connected vehicles and intelligent manufacturing [6] - The alliance is focused on accelerating the transition of research outcomes from laboratories to production lines through innovative application scenarios [7] Funding and Support - The Guangdong Intelligent Industry Fund aims to support high-tech sectors with a target size of 10 billion yuan, focusing on long-cycle, high-barrier areas such as embodied intelligence and AI chips [10] - The fund has already secured 6 billion yuan in its first phase and is designed to act as "patient capital" to support critical technology development [10] Future Outlook - In the next 3-5 years, Guangdong aims to integrate AI technology into consumer electronics, enhancing product functionality and value [11] - The alliance plans to release the "AIR Pearl River Index" and a white paper to guide industry development and avoid homogenization, focusing on high-growth sectors [11]