万兴科技集团股份有限公司 关于回购公司股份的进展公告 本公司及董事会全体成员保证信息披露的内容真实、准确、完整，没有 虚假记载、误导性陈述或重大遗漏。 万兴科技集团股份有限公司（以下简称"公司"）于 2024 年 8 月 15 日召开 第四届董事会第十九次会议，审议通过了《关于回购公司股份方案的议案》，使 用自有资金以集中竞价方式回购公司部分 A 股股份，用于员工持股计划或股权 激励计划；回购股份的资金总额不低于人民币 2,000 万元（含）且不超过人民币 4,000 万元（含），回购股份的价格不超过人民币 71.78 元/股（含），回购期限 自公司董事会审议通过回购股份方案之日起 12 个月内。具体内容详见公司于 2024 年 8 月 16 日在巨潮资讯网上披露的《关于回购公司股份方案的公告暨回购 报告书》（公告编号：2024-054）。 证券代码：300624 证券简称：万兴科技 公告编号：2025-040 证券代码：300624 证券简称：万兴科技 公告编号：2025-040 公司回购股份的时间、回购股份的价格以及集中竞价交易的委托时段等均符 合公司回购股份方案及《深圳证券交易所上市公司自律监管指引第 ...

Core Insights - The article discusses the core technologies of multimodal large models, focusing on representation learning, translation, alignment, fusion, and collaborative learning [1][2][7][11][14]. Representation Learning - Representation learning is fundamental for multimodal tasks, addressing challenges such as combining heterogeneous data and handling varying noise levels across different modalities [1]. - Prior to the advent of Transformers, different modalities required distinct representation learning models, such as CNNs for computer vision (CV) and LSTMs for natural language processing (NLP) [1]. - The emergence of Transformers has enabled the unification of multiple modalities and cross-modal tasks, leading to a surge in multimodal pre-training models post-2019 [1]. Translation - Cross-modal translation aims to map source modalities to target modalities, such as generating descriptive sentences from images or vice versa [2]. - The use of syntactic templates allows for structured predictions, where specific words are filled in based on detected attributes [2]. - Encoder-decoder architectures are employed to encode source modality data into latent features, which are then decoded to generate the target modality [2]. Alignment - Alignment is crucial in multimodal learning, focusing on establishing correspondences between different data modalities to enhance understanding of complex scenarios [7]. - Explicit alignment involves categorizing instances with multiple components and measuring similarity, utilizing both unsupervised and supervised methods [7][8]. - Implicit alignment leverages latent representations for tasks without strict alignment, improving performance in applications like visual question answering (VQA) and machine translation [8]. Fusion - Fusion combines multimodal data or features for unified analysis and decision-making, enhancing task performance by integrating information from various modalities [11]. - Early fusion merges features at the feature level, while late fusion combines outputs at the decision level, with hybrid fusion incorporating both approaches [11][12]. - The choice of fusion method depends on the task and data, with neural networks becoming a popular approach for multimodal fusion [12]. Collaborative Learning - Collaborative learning utilizes data from one modality to enhance the model of another modality, categorized into parallel, non-parallel, and hybrid methods [14][15]. - Parallel learning requires direct associations between observations from different modalities, while non-parallel learning relies on overlapping categories [15]. - Hybrid methods connect modalities through shared datasets, allowing one modality to influence the training of another, applicable across various tasks [15].

Core Insights - The multi-modal large model market in China is projected to reach 15.63 billion yuan in 2024, an increase of 6.54 billion yuan from 2023, and is expected to grow to 23.48 billion yuan in 2025, indicating strong market demand and government support [1][6][19] Multi-Modal Large Model Industry Definition and Classification - Multi-modal large models are AI systems capable of processing and understanding various data forms, including text, images, audio, and video, using deep learning technologies like the Transformer architecture [2][4] Industry Development History - The multi-modal large model industry has evolved through several stages: task-oriented phase, visual-language pre-training phase, and the current multi-modal large model phase, focusing on enhancing cross-modal understanding and generation capabilities [4] Current Industry Status - The multi-modal large model industry has gained significant attention due to its data processing capabilities and diverse applications, with a market size projected to grow substantially in the coming years [6][19] Application Scenarios - The largest application share of multi-modal large models is in the digital human sector at 24%, followed by gaming and advertising at 13% each, and smart marketing and social media at 10% each [8] Industry Value Chain - The industry value chain consists of upstream components like AI chips and hardware, midstream multi-modal large models, and downstream applications across various sectors including education, gaming, and public services [10][12] Competitive Landscape - Major players in the multi-modal large model space include institutions and companies like the Chinese Academy of Sciences, Huawei, Baidu, Tencent, and Alibaba, with various models being developed to optimize training costs and enhance capabilities [16][17] Future Development Trends - The multi-modal large model industry is expected to become more intelligent and humanized, providing richer and more personalized user experiences, with applications expanding across various fields such as finance, education, and content creation [19]

证券代码：300624 证券简称：万兴科技 公告编号：2025-038 本次回购股份资金来源为公司自有资金及回购股票专项贷款资金。本次回购 符合回购公司股份方案及相关法律法规的要求。 二、其他说明 万兴科技集团股份有限公司 关于首次回购公司股份的公告 本公司及董事会全体成员保证信息披露的内容真实、准确、完整，没有 虚假记载、误导性陈述或重大遗漏。 万兴科技集团股份有限公司（以下简称"公司"）于 2025 年 4 月 25 日召开 第四届董事会第二十四次会议审议通过了《关于回购公司股份方案的议案》，同 意使用自有资金和自筹资金通过深圳证券交易所股票交易系统以集中竞价交易 方式回购公司股份，用于员工持股计划或股权激励计划，回购股份的期限自董事 会审议通过本次回购股份方案之日起 12 个月内，回购股份的价格为不超过人民 币 89.85 元/股（含），回购的资金总额不低于人民币 2,500 万元（含），不超过 人民币 5,000 万元（含）。具体内容详见公司于 2025 年 4 月 28 日在巨潮资讯网 上披露的《关于回购公司股份方案的公告》（公告编号：2025-019）。 根据《上市公司股份回购规则》《深圳证券交易 ...

Investment Rating - The report does not provide a specific investment rating for Wondershare (300624.SZ) as it is categorized as "Not Covered" [1]. Core Insights - The management of Wondershare has a positive outlook on user base stickiness and revenue growth from entertainment applications, indicating strong user willingness to pay [7][8]. - The company is focusing on enhancing its AI-Native applications and AI agent capabilities, with plans to upgrade AI algorithms based on user acceptance [1][8]. Summary by Relevant Sections User Base and Revenue Growth - Wondershare's user base is primarily supported by professional users, showing high stickiness and an upward trend in renewal rates, reflecting strong willingness to pay [7]. - Revenue from entertainment products like SelfyzAI and Media.io studio has been increasing rapidly over the past year, alongside efforts to attract new users [7]. AI-Native Applications and Market Outlook - There is strong user interest in AI-Native applications such as SelfyzAI, and the company plans to enhance awareness of its entertainment products [8]. - A prototype AI agent has been applied to Filmora, with ongoing development to meet commercial standards [8]. Company Profile - Wondershare specializes in multimedia, data recovery, and PDF solutions, actively integrating AI technologies into its flagship products [3]. - The company is investing in R&D for video and audio algorithms to enhance its AI software functions and is collaborating with external foundation models [3].

Group 1 - The 21st China (Shenzhen) International Cultural Industries Fair showcases the integration of culture and technology, with a focus on Hunan's cultural heritage and innovative achievements [3][6] - Hunan Museum and Hongrui Cultural Group have developed AI interactive digital display cabinets that allow visitors to engage with cultural relics and learn their historical significance [3][5] - The exhibition area of Hunan covers 1,134 square meters and features the theme "Huxiang Cultural Pulse, Sci-tech Future," attracting audiences from over 110 countries and regions [3][10] Group 2 - The "Digital Han Life" project, developed by Hunan Museum in collaboration with the Malanshan Cultural Digital Innovation Center, expands the application of artifacts from the Mawangdui Han Tomb through digital databases [8] - The Mawangdui pattern digital gene library is a systematic database based on over 3,000 Han Dynasty patterns, enabling high-precision restoration and cultural interpretation [8] - The fair attracted over 6,000 exhibitors from across the country and more than 300 overseas exhibitors, promoting cultural exchange and the global outreach of Huxiang culture [10] Group 3 - Wangxing Technology showcased its new video creative software, which is the first professional video editing software adapted for HarmonyOS, featuring over 20 deep adaptations for efficient editing [12] - The software integrates elements of Huxiang culture, allowing users to create videos with traditional Chinese aesthetics and music [12] - 49 Technology presented DIY cultural creation devices, enabling consumers to design their own cultural products at home, thus expanding their global service network [14] Group 4 - Mango TV's international app has accumulated 288 million downloads, with a goal to achieve 3 million daily active users and over 1 billion in revenue by 2027 [14]

Core Insights - The article discusses the development and support of the multimodal large model industry in China, highlighting various policies and initiatives at both national and local levels aimed at enhancing AI capabilities and applications [1][4][11]. Policy Development Timeline - In 2023, local policies began to emerge, focusing on computational power to encourage the development of large model technology and innovative application scenarios, starting with Guangdong, Beijing, and Shanghai. By 2024, more regions are expected to introduce relevant policies aimed at improving administrative efficiency [1]. - By 2025, government work reports will emphasize the ongoing promotion of the "Artificial Intelligence +" initiative, with a focus on supporting the widespread application of large models [1]. National Policy Summary - The Chinese government has implemented several measures to support the AI industry, particularly multimodal large models, which are seen as crucial products within the AI sector. The State Council has identified embodied intelligence as a future industry, promoting the integration of digital technology with manufacturing and market advantages [4][5]. - Key national policies include the "Guidelines for the Development of Artificial Intelligence Industry" and the "Three-Year Action Plan for Data Elements," which aim to enhance data utilization and promote high-quality economic development through data-driven initiatives [11][13]. Local Policy Highlights - Various provinces have introduced specific policies to support the development of AI large models. For instance, Guangdong aims to develop a comprehensive technology system for large models with trillion-parameter capabilities, while Beijing targets the creation of 3-5 advanced, controllable foundational model products by the end of 2025 [13][15]. - Local initiatives also include the establishment of intelligent computing centers and the promotion of AI applications in various sectors, such as manufacturing, healthcare, and urban governance [13][14]. Key Development Directions - The article outlines that provinces like Guangdong, Beijing, and Shanghai have set ambitious goals for the development of large models, focusing on creating a robust ecosystem for AI innovation and application [15]. - The emphasis is on fostering collaboration between government, industry, and academia to drive advancements in AI technologies and their practical applications across different sectors [15].

Core Viewpoint - The 21st China (Shenzhen) International Cultural Industries Fair showcased innovative AI products from Wankey Technology, highlighting the launch of the Wankey Miaoying Harmony version, which emphasizes user-friendly video creation through AI technology [1][3][10]. Group 1: Product Highlights - Wankey Miaoying Harmony version features dual-screen multi-touch editing, allowing users to create videos without traditional input devices, making video creation accessible to everyone [3]. - The AI video creation tool, Wankey Bobo, offers a comprehensive AIGC creation experience, integrating script writing, AI translation, and video generation with a vast library of digital personas [3][10]. - Wankey SelfyzAI provides diverse AI functionalities, including dynamic photo transformation and personalized digital avatars, enhancing user creativity [3]. Group 2: Company Overview - Wankey Technology's "Tianmu" model, the first multimedia AI model in China, utilizes 1.5 billion user behavior data and 10 billion localized high-quality audio and video data to offer nearly 100 features, including AI digital personas and video stylization [10]. - The company operates in over 200 countries and regions, with a cumulative global user base exceeding 1.5 billion, positioning itself as the "Chinese version of Adobe" [10].

Core Insights - The 21st China (Shenzhen) International Cultural Industries Fair showcases the integration of culture and technology, highlighting Hunan's cultural heritage and innovative practices [1][3][5] Group 1: Event Overview - The event runs from May 22 to May 26, featuring Hunan's exhibition themed "Huxiang Culture and Sci-tech Future" [1] - The exhibition includes various sections such as the main image area, comprehensive exhibition of cultural and technological integration, and a special area for audio-visual industries [1] Group 2: Technological Innovations - Wanxing Technology launched the Wanxing Miaoying Hongmeng version, a video editing software that offers over 20 features optimized for Huawei's HarmonyOS [1][9] - The Ma Lan Mountain Audio-Video Laboratory presented eight proprietary audio-visual technologies, including a dual-lens AI live streaming camera and a wireless sound system that enhances home audio experiences [7][9] Group 3: Cultural Exhibits - The Hunan Broadcasting System showcased a digital restoration of cultural relics, integrating technology with storytelling to present the history behind artifacts [3][5] - Interactive digital display cabinets developed by Hunan Museum and Hongrui Cultural Heritage Group allow visitors to engage with artifacts, providing historical context and narratives [5] Group 4: Future Prospects - The Ma Lan Mountain Micro Short Drama Intelligent Translation Center aims to produce over 10,000 high-quality translated micro short dramas by 2025, utilizing advanced AI translation tools [9]

Group 1 - The Sora (Wensheng Video) concept increased by 0.31%, ranking third in the concept sector, with 11 stocks rising, including Kunlun Wanwei which hit a 20% limit up [1][2] - The leading gainers in the Sora concept include Lingyun Guang with a 5.74% increase, Yinsai Group with a 4.95% increase, and Wanxing Technology with a 2.90% increase [1][2] - The main losers in the sector were Aorui De, *ST Tianze, and Jishi Media, which fell by 2.75%, 2.73%, and 2.65% respectively [1][2] Group 2 - The Sora concept saw a net inflow of 1.348 billion yuan, with 13 stocks receiving net inflows, and 5 stocks exceeding 30 million yuan in net inflow [2][3] - Kunlun Wanwei led the net inflow with 1.278 billion yuan, followed by Lingyun Guang, Yinsai Group, and Jiecheng Shares with net inflows of 65.303 million yuan, 46.0198 million yuan, and 38.3301 million yuan respectively [2][3] Group 3 - In terms of net inflow ratios, Kunlun Wanwei, Zhidema, and Jiecheng Shares had the highest ratios at 23.76%, 12.77%, and 10.63% respectively [3] - The trading volume and turnover rates for the leading stocks in the Sora concept were significant, with Kunlun Wanwei having a turnover rate of 12.32% and a daily increase of 20.01% [3]