丨证券研究报告丨 公司研究丨点评报告丨三人行（605168.SH） [Table_Title] 汽车客户投放减少拖累营收，新业务落地进展积 极 高超 朱子愉 请阅读最后评级说明和重要声明 %% %% %% %% 报告要点 [Table_Summary] 公司公布 2025 半年报：2025 年上半年，公司实现营业收入 16.57 亿元，同比下滑 13.36%， 实现归母净利润 1.44 亿元，同比增长 10.83%；其中 25Q2 公司实现营业收入 8.40 亿元，同比 下滑 13.93%，实现归母净利润 0.71 亿元，同比下滑 13.15%。公司紧扣广告主业，积极拓展 增量，同时积极围绕新质生产力开展新业务，体育彩票新业务有望迎来突破性进展。 分析师及联系人 [Table_Author] SAC：S0490516080001 SFC：BUX177 公司公布 2025 半年报：2025 年上半年，公司实现营业收入 16.57 亿元，同比下滑 13.36%， 实现归母净利润 1.44 亿元，同比增长 10.83%，实现扣非归母净利润 1.27 亿元，同比增长 4.56%；其中 25Q2 公司实现营业收入 8. ...

生成式AI 一、 Meta重组AI部门，将超级智能实验室拆分为四个团队 https://mp.weixin.qq.com/s/GzaD8SM82Sy5IdQ6wBjl2g 二、 实测最新DeepSeek V3.1 Base，不止拓展上下文长度 1. DeepSeek V3.1相比V3不仅拓展上下文长度至128k，在编程表现、创意写作、翻译水平和回答语气等方面均有明 显提升； 2. 实测表明，V3.1在代码能力方面更全面周到，考虑了更多可能性并主动提供使用说明，支持更激进的压缩策略； 3. V3.1在Reddit测试中获得71.6%分数，成为非推理模型SOTA，比Claude Opus 4高1%但价格便宜68倍。 https://mp.weixin.qq.com/s/x0X481MgH_8ujjB0_XL4SQ 三、 智谱发布AutoGLM 2.0，为Agent配置云手机/云电脑 1. Meta将超级智能实验室拆分为TBD Lab（研究新版Llama）、FAIR（长期研究）、产品应用团队和基础设施四个 部门； 2. 新团队已讨论将Meta下一代AI模型改为闭源模式，可能放弃Llama 4转而从头开发新模型，动摇 ...

Core Viewpoint - The company Electric Science and Technology Network Security (电科网安) has developed a key product, the "Sensitive Data Discovery and Classification System," which utilizes multi-modal AI for precise identification of unstructured data [2] Group 1: Product Overview - The "Sensitive Data Discovery and Classification System" is one of the earliest products in the data security field for the company [2] - The product team has successfully tackled several core technologies, including automated asset discovery and AI-based automated data classification and grading [2] - The system features sensitive data flow monitoring and industry classification and grading templates [2] Group 2: Technology and Capabilities - The system employs multi-modal AI for precise identification of sensitive data within unstructured data, integrating image recognition and natural language processing technologies [2] - It supports various common types of unstructured data, including office documents, with industry-leading accuracy in identifying sensitive data [2]

Core Viewpoint - The article discusses the journey of Binson, the founder of "Xinying Suixing," and his innovative AI product "Doudou Game Partner," which aims to provide companionship in gaming and beyond, with a focus on user engagement and emotional connection [5][6][20]. Company Overview - Binson, a former executive at major tech companies, founded "Xinying Suixing" in 2023 after a life-changing experience [6]. - The company completed a multi-million dollar A+ funding round at the end of last year, with investors including Jiuhe Venture Capital and Xinying Capital [7]. Product Development - The first version of "Doudou Game Partner" was launched after overcoming initial model limitations, evolving into a Companion AI capable of real-time interaction during gameplay [6][7]. - The latest update (version 1.0) enhances the product's ability to understand and interact with users in various gaming scenarios [7][8]. User Engagement - "Doudou Game Partner" has achieved 8 million registered users and over 2 million monthly active users, indicating strong market traction [6]. - The typical user demographic is young adults aged 18-25, who seek companionship while gaming [12]. Usage Scenarios - Users commonly engage with "Doudou" for in-game discussions, strategy assistance, and even watching shows together [11][12]. - The product aims to fulfill emotional needs by providing a non-judgmental companion for users during their gaming experiences [20]. Monetization Strategy - Current monetization includes in-game purchases and subscription models, with future plans to incorporate B2B advertising and e-commerce recommendations [14][15]. - The company anticipates a balanced revenue stream from both consumer and business segments [15]. Future Aspirations - Binson expresses a desire to enhance the AI's capabilities for understanding complex gaming events and providing detailed feedback to users [16][17]. - The long-term vision includes extending the emotional connection beyond gaming into users' daily lives [20].

Core Viewpoint - Kunlun Wanwei has launched six new models in one week, showcasing its advancements in multimodal AI applications, including video generation, world models, and AI music creation, indicating a strategic push in the AI sector [2][5][63]. Group 1: Model Launches - The company released the SkyReels-A3 model, designed for digital human live-streaming, which can generate realistic videos driven by audio input, enhancing the e-commerce landscape [9][10][16]. - Matrix-Game 2.0, an upgraded interactive world model, was introduced, boasting real-time generation and long-sequence capabilities, positioning it as a competitor to Google's Genie 3 [19][20][22]. - The Matrix-3D model was launched, integrating panoramic video generation and 3D reconstruction, breaking barriers between content generation and interaction [25][27]. - Skywork UniPic 2.0 was unveiled as a unified multimodal model capable of image understanding, generation, and editing, demonstrating a new training paradigm that reduces hardware requirements [29][31][33]. - The Skywork Deep Research Agent v2 was released, enhancing multimodal capabilities for deep research and content generation [37][38]. - Mureka V7.5, a music generation model, was launched, focusing on Chinese music, showcasing significant improvements in emotional expression and musicality [53][54][56]. Group 2: Strategic Insights - Kunlun Wanwei's strategy emphasizes vertical integration in AI, focusing on high-frequency application scenarios rather than general-purpose agents, which is seen as a more viable approach for future development [70][72][76]. - The company has committed substantial resources to R&D, with a projected R&D expenditure of 1.54 billion yuan in 2024, reflecting a 59.5% year-on-year increase, and a workforce of 1,554 dedicated to AI research [73][74]. - The open-source approach adopted by Kunlun Wanwei has positioned it as a leader in the AI ecosystem, contributing to its recognition as one of the "Top 16 AI Open Source Companies in China" [5][78].

Core Insights - The article highlights the development of the Zengcheng Low Carbon Headquarters Park as a hub for technology-driven enterprises, focusing on fostering innovation and providing financial support to small and medium-sized enterprises (SMEs) [1][9] Group 1: Financial Support and Ecosystem - The park has attracted over 20 financial institutions, including commercial banks and investment funds, creating a comprehensive capital empowerment system for enterprises from startup to maturity, with over 2 billion yuan in financing provided in 2024 alone [1][3] - Various financing products are offered, including short-term credit products tailored for SMEs, with online approval processes allowing loans to be disbursed within three days [2][3] - The park has issued over 20 billion yuan in loans to resident companies, supporting numerous enterprises with financing exceeding 10 million yuan each [3] Group 2: Innovation and Entrepreneurship - The park has hosted the China Innovation and Entrepreneurship Competition for three consecutive years, serving as a significant platform for resource aggregation and providing financing connections for winning projects [7][8] - The integration of advanced technologies such as digital twins and AI in companies like Guangdong Yuan Neng Xing Tai demonstrates the park's focus on facilitating the commercialization of innovative research [7][8] Group 3: Comprehensive Services - The park has established six public service platforms to cater to the diverse needs of enterprises at different stages, including talent introduction, entrepreneurial incubation, and technology-market connections [9] - A collaborative initiative with various financial and securities institutions aims to create a nurturing environment for companies preparing for public listings, with 14 companies currently in the pipeline for potential listing [9][10]

截至2025年8月15日收盘，云鼎科技股价报13.12元，较前一交易日上涨0.37元，涨幅2.90%。当日开盘 价为12.76元，最高触及13.19元，最低下探至12.60元，成交量为47.75万手，成交额达6.20亿元。 资金流向方面，8月15日主力资金净流入4944.07万元，占流通市值的0.89%。近五个交易日主力资金累 计净流出191.71万元。 风险提示：股市有风险，投资需谨慎。 云鼎科技主营业务涵盖互联网服务、多模态AI、数据要素等领域，公司注册地位于山东。 消息面上，云鼎科技在互动平台表示，公司2025年半年度报告将于8月27日披露。 ...

Core Viewpoint - The company Zhejiang Zhongyin Fashion Co., Ltd. is experiencing fluctuations in stock performance and is involved in various innovative technologies related to virtual digital humans and AI, while also benefiting from the depreciation of the RMB. Group 1: Company Performance - On August 14, Zhongyin Fashion's stock fell by 2.77%, with a trading volume of 108 million yuan and a market capitalization of 4.214 billion yuan [1] - The company reported a revenue of 78.9853 million yuan for the first quarter of 2025, representing a year-on-year growth of 4.96%, while the net profit attributable to the parent company was -2.6389 million yuan [7] - The company has distributed a total of 83.3324 million yuan in dividends since its A-share listing, with 59.3324 million yuan distributed over the past three years [8] Group 2: Technological Advancements - The company has made significant advancements in virtual digital human technology, with multiple international-leading technologies in 3D digital human generation and AIGC+3D digital human AI cross-modal real-time interaction [2][3] - The first-generation digital human product "Chuangshiyuan" supports AIGC multi-modal content generation, allowing for quick recognition and intelligent video generation from various formats [2] Group 3: Revenue Composition and Market Position - The company's revenue composition includes 80.77% from supply chain integration, 10.62% from design services, 3.56% from brand operation, 1.95% from shoe production, and 1.59% from cultural tourism services [7] - As of the latest report, the company's overseas revenue accounted for 83.07%, benefiting from the depreciation of the RMB [3]

Core Viewpoint - The article emphasizes the rapid development and open-sourcing of domestic AI models in China, particularly highlighting the advancements made by Kunlun Wanwei in the field of multi-modal AI and intelligent agents [1][47]. Group 1: Open-source Models and Developments - In July, the Chinese AI community saw an impressive total of 33 open-source models released, with major players like Kunlun Wanwei, Alibaba, and Tencent participating [1]. - In August, Kunlun Wanwei continued to release significant models, including the second-generation reward model Skywork-Reward-V2 and the multi-modal understanding model Skywork-R1V3 [1]. - Kunlun Wanwei launched a week-long technology release event, showcasing various models across multi-modal AI applications [1]. Group 2: Skywork Deep Research Agent - On August 14, Kunlun Wanwei released the upgraded version of its Skywork Deep Research Agent, enhancing its capabilities in multi-modal information retrieval and generation [3]. - The Skywork Deep Research Agent achieved a remarkable accuracy of 27.8% in conventional reasoning mode and 38.7% in its proprietary "parallel thinking" mode, setting a new industry SOTA record [4]. - The agent also excelled in the GAIA benchmark test, surpassing all competitors in complex task performance [6]. Group 3: Multi-modal Capabilities - Kunlun Wanwei's agent integrates multi-modal retrieval and understanding, allowing it to process images and charts, thus enhancing the completeness and accuracy of research reports [12]. - The agent can generate detailed reports with rich visual content, including graphs and charts, while ensuring that all data sources are cited [21][22]. - The system employs advanced technologies such as MM-Crawler for efficient data collection and multi-agent architecture for task execution [29][30]. Group 4: Technological Innovations - The Skywork Deep Research Agent V2 incorporates several key enhancements, including high-quality data synthesis, end-to-end reinforcement learning, and efficient parallel reasoning [40]. - The agent's architecture allows for dynamic task management and collaboration among multiple agents, improving adaptability and efficiency [44]. - Innovations in data quality standards and complex problem-solving strategies have been implemented to enhance the agent's learning and reasoning capabilities [41][42]. Group 5: Industry Trends and Future Outlook - The article notes a shift in the AI industry focus from developing singular powerful models to open-source collaboration and practical application deployment [47]. - Companies that can effectively build comprehensive toolchains and application ecosystems on top of open-source models are likely to gain a competitive edge in the AI landscape [49]. - Kunlun Wanwei's recent developments signal its commitment to advancing multi-modal AI and establishing a strong position in the global AI competition [50].

Core Viewpoint - The article discusses the advancements in AI memory, particularly focusing on visual memory as a crucial component for achieving Artificial General Intelligence (AGI). Memories.ai aims to create a foundational visual memory layer that allows AI to "see and remember" the world, overcoming the limitations of current AI systems that primarily rely on text-based memory [2][8][9]. Group 1: Visual Memory Technology and AI Applications - Memories.ai is developing a Large Visual Memory Model (LVMM) that is inspired by human memory systems, aiming to enable AI to process and retain vast amounts of visual data [22][25]. - The distinction between text memory and visual memory is emphasized, with the former being more about context engineering rather than true memory, while visual memory aims to replicate human-like understanding and retention of information [13][14]. - The company is positioning itself as a B2B infrastructure provider, enabling other AI companies and traditional industries like security, media, and marketing to leverage its visual memory technology [31][34]. Group 2: Technical Challenges and Infrastructure - The LVMM system is designed to handle the unique challenges of video data, such as high volume and low signal-to-noise ratio, through a complex architecture that includes compression, indexing, and retrieval mechanisms [22][27]. - The ability to manage petabyte-scale infrastructure is highlighted as a key competitive advantage for building a global visual memory system [28][30]. - The company’s infrastructure is capable of supporting a vast database for efficient querying and retrieval, which is essential for scaling its visual memory capabilities [28][30]. Group 3: Industry Applications and Future Directions - The technology has potential applications in various sectors, including real-time security detection, media asset management, and video marketing, with ongoing collaborations with major companies in these fields [34][35]. - The future vision includes developing AI assistants and humanoid robots that possess visual memory, enabling them to interact with users in a more personalized manner [39][41]. - The company is also exploring partnerships with AI hardware firms to enhance the capabilities of its visual memory technology in consumer applications [36][41].