香港交易及結算所有限公司及香港聯合交易所有限公司對本公告的內容概不負責，對其準確性 或完整性亦不發表任何聲明，並明確表示，概不對因本公告全部或任何部份內容而產生或因倚 賴該等內容而引致的任何損失承擔任何責任。 董事會召開日期 商湯集團股份有限公司（「本公司」，連同其附屬公司及合併聯屬實體統稱「本集 團」）董事會（「董事會」）茲通告，本公司謹定於2025年8月28日（星期四）舉行董事 會會議，以（其中包括）考慮及批准本集團截至2025年6月30日止六個月的中期業 績及其刊發，及考慮派發中期股息（如有）。 承董事會命 SenseTime Group Inc. 商湯集團股份有限公司 執行主席 首席執行官 徐立博士 香港，2025年8月12日 於本公告日期，執行董事為徐立博士、王曉剛博士、林達華博士、楊帆先生及 王征先生；非執行董事為范瑗瑗女士；以及獨立非執行董事為薛瀾教授、林怡仲 先生及邱達根先生。 （於開曼群島註冊成立以不同投票權控制的有限公司） （股份代號：0020（港幣櫃台）及80020（人民幣櫃台）） ...

Core Viewpoint - The article emphasizes the significance of "multimodal intelligence" as a key trend in the development of large models, particularly highlighted during the WAIC 2025 conference, where SenseTime introduced its commercial-grade multimodal model, "Riri Xin 6.5" [1][2]. Group 1: Importance of Multimodal Intelligence - Multimodal intelligence is deemed essential for achieving Artificial General Intelligence (AGI) as it allows AI to interact with the world in a more human-like manner, processing various forms of information such as images, sounds, and text [7][8]. - The article discusses the limitations of traditional language models that rely solely on text data, arguing that true AGI requires the ability to understand and integrate multiple modalities [8]. Group 2: Technical Pathways to Multimodal Models - SenseTime has identified two primary technical pathways for developing multimodal models: Adapter-based Training and Native Training. The latter is preferred as it allows for a more integrated understanding of different modalities from the outset [11][12]. - The company has committed significant computational resources to establish a "native multimodal" approach, moving away from a dual-track system of language and image models [10][12]. Group 3: Evolutionary Path of Multimodal Intelligence - SenseTime outlines a "four-breakthrough" framework for the evolution of AI capabilities, which includes advancements in sequence modeling, multimodal understanding, multimodal reasoning, and interaction with the physical world [13][22]. - The introduction of "image-text intertwined reasoning" is a key innovation that allows models to generate and manipulate images during the reasoning process, enhancing their cognitive capabilities [16][18]. Group 4: Data Challenges and Solutions - The article highlights the challenges of acquiring high-quality image-text pairs for training multimodal models, noting that SenseTime has developed automated pipelines to generate these pairs at scale [26][27]. - SenseTime employs a rigorous "continuation validation" mechanism to ensure data quality, only allowing data that demonstrates performance improvement to be used in training [28][29]. Group 5: Model Architecture and Efficiency - The focus on efficiency over sheer size in model architecture is emphasized, with SenseTime optimizing its model to achieve over three times the efficiency while maintaining performance [38][39]. - The company believes that future model development will prioritize performance-cost ratios rather than simply increasing parameter sizes [39]. Group 6: Organizational and Strategic Insights - SenseTime's success is attributed to its strong technical foundation in computer vision, which has provided deep insights into the value of multimodal capabilities [40]. - The company has restructured its research organization to enhance resource allocation and foster innovation, ensuring a focus on high-impact projects [41]. Group 7: Long-term Vision and Integration of Technology and Business - The article concludes that the path to AGI is a long-term endeavor that requires a symbiotic relationship between technological ideals and commercial viability [42][43]. - SenseTime aims to create a virtuous cycle between foundational infrastructure, model development, and application, ensuring that real-world challenges inform research directions [43].

Group 1: Investment Trends in Chinese High-Tech Sector - International long-term capital is increasingly interested in investing in China's high-tech sector, with 59% of sovereign funds planning to increase their allocation to Chinese assets over the next five years [1][2] - Temasek, Singapore's sovereign wealth fund, has an investment exposure to Chinese assets of 18%, significantly higher than the 2%-3% allocation in the MSCI global index [1] - Hong Kong's IPO and secondary financing raised over 180 billion HKD in the first half of the year, a 62% year-on-year increase, indicating strong international participation in Chinese tech IPOs [1] Group 2: AI Sector Investment Insights - International long-term capital views leading Chinese AI companies as having strong technological depth and global reach, creating significant investment opportunities [2][5] - The average price-to-sales (P/S) ratio for leading Chinese AI companies is approximately 15, compared to 42 for top US AI firms, indicating that Chinese AI stocks are perceived as undervalued [6] - The rapid improvement in AI capabilities in China, particularly in large-scale multi-task language understanding, is attracting renewed interest from global investors [4] Group 3: Investment Strategies and Preferences - Many international long-term investors prefer participating in Hong Kong's IPOs and private placements rather than venture capital or private equity investments due to smoother exit strategies and better risk management [2][3] - The regulatory environment in Hong Kong, which aligns with international standards while understanding local market rules, makes it an attractive hub for international capital [7] - Investment firms are focusing on the stock selection capabilities of asset managers, as long-term returns are primarily driven by stock selection rather than market timing [8]

Core Viewpoint - The article discusses the emergence of embodied intelligence and the significance of the "world model" as a core component in advancing AI towards human-like intelligence, highlighting the competitive landscape in the AI industry as it evolves towards embodied intelligence [1][2]. Industry Developments - Major companies like Google, Huawei, and ByteDance are launching various embodied intelligence platforms and models, indicating a rapid evolution in this field [3]. - SenseTime, leveraging its expertise in computer vision and multi-modal large models, aims to empower the industry through its "Wuneng" embodied intelligence platform, which integrates years of technological accumulation [3][5]. Technical Challenges - The industry faces challenges such as data scarcity, difficulty in large-scale production, and the need for generalization in embodied intelligence applications [5][13]. - The reliance on computer vision expertise is seen as a potential solution to enhance the learning of world models and improve the capabilities of embodied intelligence [14]. World Model Significance - The world model is recognized as a crucial element for predicting and planning in autonomous systems, enabling robots to interact intelligently with their environments [12][17]. - SenseTime's "Kaigu" world model is designed to provide extensive data and facilitate simulation-based learning, significantly reducing data collection costs [17][20]. Platform Features - The "Wuneng" platform offers a comprehensive approach by combining first-person and third-person perspectives for robot learning, enhancing the understanding of robot behavior [27][29]. - The platform aims to address the data challenges in the industry by providing synthetic data and facilitating the development of various robotic applications [26][31]. Future Implications - As embodied intelligence matures, it is expected to transform human-robot interactions and create new social networks involving robots, enhancing their roles in daily life [36][37]. - The integration of embodied intelligence into common environments like homes and workplaces is anticipated to unlock significant value and functionality [39].

Group 1 - The core viewpoint is that major AI models like Baidu's Wenxin Yiyan and SenseTime's SenseChat are now open to the public, marking a significant step in the maturity of generative AI technology and its regulatory framework [1][2] - Baidu has launched the Wenxin Yiyan app for users to download and experience, while enterprise users can access its capabilities through Baidu's intelligent cloud platform [1] - SenseTime has established deep collaborations with over 500 clients across various industries, providing flexible API interfaces and services for generative AI applications [1] Group 2 - The launch of domestic large models is expected to create multiple opportunities for companies involved in their development, as well as for data providers and algorithm optimization firms [2] - Several listed companies have reported advancements in large models, with Oriental Guoxin accelerating the deployment of BonGPT and Kunlun Wanwei's "Tiangong" model showing strong natural language processing capabilities [2] - The rapid promotion and application of large models by tech giants like Baidu and SenseTime indicate a recognition of the commercial value of generative AI, which is likely to stimulate competition and drive technological innovation [2]

Summary of Key Points Core Viewpoint - The report highlights the top short-selling stocks in the market, indicating significant investor sentiment against these companies, with notable short-selling ratios and amounts for specific stocks [1][2]. Group 1: Top Short-Selling Ratios - Anta Sports (82020) has the highest short-selling ratio at 100.00% with a short-selling amount of 97,800 CNY [2]. - China National Offshore Oil Corporation (80883) follows with a short-selling ratio of 90.59% and a short-selling amount of 1,305,000 CNY [2]. - Tencent Holdings (00700) has a short-selling ratio of 87.99% with a short-selling amount of 1,125,000 CNY [2]. Group 2: Top Short-Selling Amounts - Meituan (03690) leads in short-selling amount with 1.136 billion CNY and a short-selling ratio of 23.98% [2]. - Xiaomi Group (01810) has a short-selling amount of 1.04 billion CNY and a short-selling ratio of 16.20% [2]. - Tencent Holdings (00700) also appears in this category with a short-selling amount of 1.007 billion CNY and a short-selling ratio of 15.29% [2]. Group 3: Top Short-Selling Deviation Values - Tencent Holdings (80700) has the highest deviation value at 45.57%, indicating a significant difference from its average short-selling ratio over the past 30 days [2]. - China National Offshore Oil Corporation (80883) has a deviation value of 42.14% [2]. - China State Construction International (03311) shows a deviation value of 34.50% [2].

Group 1 - The core viewpoint is that Chinese AI companies are becoming attractive to international long-term capital, indicating a historic revaluation of Chinese assets [1][2] - Chinese AI enterprises are narrowing the technological gap with Western giants, with performance differences in top models decreasing from 20% in 2023 to 0.3% by the end of 2024 [1] - China has unique advantages in AI, including a rich application landscape and supportive policies, making it the world's largest AI testing ground [1] Group 2 - Despite a strong performance in the A-share market, Chinese AI companies still have potential for valuation increases, driven by breakthroughs in technology and market opportunities [2] - The average price-to-sales ratio for U.S. AI companies is 42 times, while comparable Chinese companies average around 15 times, indicating that Chinese AI firms are undervalued [2] - The value generated by AI is beginning to show, particularly in programming and office applications, suggesting that a commercial turning point for the industry is imminent [2] Group 3 - AI research and development require patient capital, and long-term funds like Infini Capital align well with the commercialization cycle of AI [3] - The focus is on technological positioning in frontier fields rather than short-term stock price fluctuations, with confidence that leading Chinese AI firms will drive the next wave of technological revolution [3]

Group 1 - SenseTime, in collaboration with various domestic partners, launched the SenseTime Large Device Computing Mall, marking a significant step in promoting domestic ecological construction and creating a one-stop AI infrastructure [1] - The SenseTime Large Device Computing Mall aims to lower the barriers for AI applications, accelerate the deployment of large models across industries, and provide users with flexible and autonomous domestic computing options, effectively breaking foreign technology barriers [1] - The computing mall offers efficient, stable, and cost-effective heterogeneous computing infrastructure, high-quality datasets, and MLOps tools for model fine-tuning, significantly enhancing development efficiency [1] Group 2 - SenseTime leverages its expertise in embodied intelligence, AIGC, AI4S, and industrial intelligence to apply the combination of "computing power + data + algorithms" to address specific business pain points, acting as an accelerator for state-owned enterprises' transition to intelligence [2] - The collaboration with Xiaomi's AI glasses integrates SenseTime's "Riri New" model interaction platform, enhancing user experience through real-time audio and video interaction capabilities, providing a reference for the next generation of smart hardware interaction paradigms [2] - SenseTime plans to deepen technological breakthroughs in real-time audio and video interaction and multimodal understanding, aiming to empower a broader range of application scenarios and support solutions from personal efficiency to industry-wide applications [3]

Group 1 - The article discusses a controversial academic paper claiming that Hammurabi and Shang Tang are the same person, which has sparked public outrage and skepticism [1][2] - The authors of the paper have been criticized for their selective use of evidence and for violating basic principles of historical linguistics, as they ignore the significant differences between Babylonian and ancient Chinese languages [1][2] - The paper's conclusions are seen as a threat to cultural confidence, as they blur the lines between Eastern and Western civilizations and could mislead future generations about historical continuity [2] Group 2 - To prevent such academic controversies, the article emphasizes the need for multiple layers of academic standards and oversight, including stricter peer review processes and accountability for academic misconduct [3] - It calls for universities to strengthen their supervision of research integrity and establish lifelong accountability mechanisms for academic misconduct [3] - The article advocates for scholars to regain respect for knowledge and adhere to rigorous academic standards to ensure the longevity and integrity of academic contributions [3]

Market Performance - As of August 8, 2025, the CSI Hong Kong Internet Index (931637) decreased by 1.33% with mixed performance among constituent stocks [3] - Leading gainers included Dongfang Zhenxuan (01797) up 7.38%, and Lion Group Holdings (02562) up 3.19%, while leading decliners were Qutoutiao (00917) down 6.40% and Kingsoft Cloud (03896) down 5.10% [3] - The Hong Kong Internet ETF (159568) fell by 1.10%, currently priced at 1.8 yuan, but has seen a 0.55% increase over the past week [3] Liquidity and Trading Activity - The Hong Kong Internet ETF had a turnover rate of 19.46% with a trading volume of 62.965 million yuan, indicating active market participation [3] - Over the past year, the average daily trading volume of the Hong Kong Internet ETF was 187 million yuan [3] User Engagement and Ecosystem - Bilibili's user ecosystem is thriving, with 368 million monthly active users and 107 million daily active users as of Q1 2025 [3] - The platform has 32 million monthly paying users and an average daily usage time of 108 minutes, with a 70% penetration rate among Gen Z users [3] ETF Performance Metrics - As of August 7, 2025, the Hong Kong Internet ETF has seen a net value increase of 72.24% over the past year, ranking 138 out of 2949 in the index fund category [5] - The ETF's highest monthly return since inception was 30.31%, with a longest consecutive monthly gain of 3 months and an average monthly return of 8.62% [5] Risk and Fee Structure - The management fee for the Hong Kong Internet ETF is 0.50% and the custody fee is 0.10%, which are among the lowest in comparable funds [7] - The ETF's tracking error over the past three months was 0.050%, indicating high tracking precision compared to similar funds [7] Valuation Insights - The latest price-to-earnings ratio (PE-TTM) for the index tracked by the Hong Kong Internet ETF is 23.8, which is below the 80.03% historical average over the past five years, suggesting a low valuation [7] - The CSI Hong Kong Internet Index consists of 30 publicly listed companies related to internet businesses, reflecting the overall performance of internet-themed stocks within the Hong Kong Stock Connect [7][8]