点击注册小程序 查看完整报告 特别申明： 本订阅号中所涉及的证券研究信息由光大证券研究所编写，仅面向光大证券专业投资者客户，用作新媒体形势下研究 信息和研究观点的沟通交流。非光大证券专业投资者客户，请勿订阅、接收或使用本订阅号中的任何信息。本订阅号 难以设置访问权限，若给您造成不便，敬请谅解。光大证券研究所不会因关注、收到或阅读本订阅号推送内容而视相 关人员为光大证券的客户。 点评： 快手618期间泛货架增长较快，为广告商业化提供流量基础 1）电商整体GMV符合预期，泛货架增速快于大盘。根据快手披露，618期间商品卡GMV同比增长超53%，搜 索GMV同比增长超143%，挂车短视频GMV同比增长超29%。泛货架作为区别于直播间的内容消费场域，已逐 步成为用户浏览、发现与购买的重要通路，内容消费与商品转化之间的路径更加清晰。2）泛货架具备进一步 广告商业化潜力，有望在25H2带来新增量。快手早期重点在于通过流量分发、供应链扶持及智能化工具建 设，引导生态内商家提升投放效率。随着中小商家逐步成熟，平台有望推动部分流量位产品化，并通过智能化 工具引导商家开展自主投放，逐步建立面向泛货架场景的广告转化链路。 全新端到 ...

凤凰网科技讯 7月1日消息，淘天集团在昨日"硬核少年技术节4.0"活动中正式发布其自主研发的百亿参 数推荐大模型RecGPT，标志着淘宝首页"猜你喜欢"功能迎来基于生成式推荐（AIGR）技术的全面革 新。 技术创新的另一亮点在于个性化推荐理由的自动生成。升级后的推荐信息流为每件商品配备了定制化的 推荐文案，如为热门玩具标注"新晋顶流不来看看吗？"，为除湿机标注地域化提醒"杭州梅雨季防潮神 器"等，提升了用户与推荐内容的互动体验。 此次推荐系统升级是淘天集团AIGX技术体系的重要应用成果。该技术体系已构建起覆盖索引、推荐、 出价、拍卖、创意、数据等电商全业务场景的AI解决方案矩阵，并在淘宝天猫多个业务线实现规模化 部署。 据官方披露的测试数据，搭载RecGPT的推荐系统已实现用户点击量两位数增长，同时用户加购行为和 页面停留时长均提升超过5%。这一技术升级代表了电商平台在个性化推荐领域的重要进展。 RecGPT以淘宝星辰LLM大模型为技术底座，通过对平台内用户历史行为数据的强化学习训练，显著增 强了电商场景下的推理分析能力。该模型能够深度解析用户在淘宝平台超过十年的消费轨迹，并运用多 模态认知技术整合数亿级商品 ...

Group 1 - The core announcement is the launch of RecGPT, a self-developed recommendation model with 100 billion parameters, which will enhance the "You Might Also Like" feature on the mobile Taobao homepage using AIGR (Generative Recommendation) technology [1] - Testing data indicates that the recommendation information flow powered by RecGPT has achieved a double-digit increase in user click-through rates, with over 5% improvements in both user add-to-cart actions and time spent on the platform [1] - RecGPT is built on the Taobao Star LLM model and utilizes reinforcement learning on user behavior data from over ten years, enhancing its reasoning capabilities in the e-commerce sector [1] Group 2 - The upgrade of the recommendation information flow represents the practical application of Taobao's AIGR technology, which is part of the broader AIGX technology system that encompasses various aspects of e-commerce operations [2] - The AIGX system includes components for indexing (AIGI), recommendation (AIGR), bidding (AIGB), auctioning (AIGA), creative generation (AIGC), and data management (AIGD), covering all necessary scenarios for e-commerce business operations [2]

有，你别说还真有。 那就是 大模型变现 。而且更细分的赛道已经很明确了—— 这不最近硅谷大厂都盯上了用 AI打广告 这门生意。 ChatGPT聊着聊着开始带货： 谷歌劈柴哥在IO大会宣布要用AI将内容和广告深度融合。Meta已经披露了实打实的数据，2024第四季度广告营收 增长21% ，都是得益于AI 的优化。 生成式AI一来，打广告的姿势变了，商业模式底层技术的探索空间，空前巨大。 普通人有机会吗？有，而且是专门面向 在校学生 的那种。 明敏 发自 凹非寺 量子位 | 公众号 QbitAI 就说当今之势，还有比搞大模型 更有前途 的吗？ 不仅有业内资深专家指导、接触实际工业数据，从小白直接变成领域内小专家，还能有奖金以及直通offer。 用大模型打广告搞钱，有啥机遇？ 用大模型搞钱姿势千千万，为啥生成式AI+广告这条路值得关注？ 最首要的，有人已经赚到钱了，实打实的营收增长正在发生。 Meta的2024年Q4财报数据显示， 广告收入占整体营收的96.7%，约468亿美元，同比增长21% 。 背后核心驱动因素是 AI 。 2024年12月，Meta官方披露了与英伟达合作的广告投放系统Andromeda。这是一 ...

Group 1: Qunar Travel - Qunar Travel held its first Customer Service Festival, introducing the "NICE" service concept focusing on nimbleness, innovation, empathy, and expertise [2] - The company announced new benefits for customer service staff, including the option to work from home for up to four days a week and an annual travel fund of 1200 RMB [2] - Qunar aims to enhance its customer service training system and leverage AI to help young staff become industry experts more quickly [2] Group 2: Midea Group - Midea Group established a new e-commerce company in Guangzhou with a registered capital of 10 million RMB, focusing on AI hardware sales and home appliance installation services [3] - The new company is jointly owned by Midea Group and its subsidiary, Foshan Midea Air Conditioning Industrial Investment Co., Ltd [3] Group 3: Tencent - Tencent launched a generative recommendation algorithm competition with a prize pool of several million and job offers to attract global talent [4] - This competition is part of Tencent's largest recruitment initiative and focuses on cutting-edge AI technology [4] Group 4: Ele.me - Ele.me announced an investment of over 1 billion RMB to enhance its quality takeout services through the "优店腾跃计划" [5] - The initiative includes offering new stores up to three months of commission-free service and increased exposure through Alibaba's ecosystem [5] Group 5: Guangxi - Guangxi has initiated the construction of an AI open innovation platform, which includes establishing new R&D institutions and joint innovation centers [6] - The platform aims to collaborate with ASEAN countries in AI research and development [6] Group 6: Baidu - Baidu's AIDU plan represents its largest recruitment drive for top AI talent, expanding job openings by over 60% compared to last year [7][8] - The recruitment covers 23 core business areas and 11 research directions, focusing on advanced AI technologies [7] Group 7: Nezha Auto - Nezha Auto's parent company has entered a restructuring process to resolve its debt crisis and optimize management [9] - The restructuring will involve bringing in a new CEO with international automotive experience and focusing on production recovery and market expansion [9] Group 8: National Bureau of Statistics - The unemployment rate for young people in China has decreased for three consecutive months, with the urban survey unemployment rate at 5% in May [10] - The growth in the wholesale and retail sectors, as well as accommodation and catering industries, has contributed to employment stability [10] Group 9: Zhilian Recruitment - The demand for jobs in the humanoid robot sector has surged, with job postings increasing by 409% year-on-year [11] - Technical positions dominate the recruitment landscape, accounting for 62% of job postings and 71% of job seekers in the robotics industry [11] Group 10: Inner Mongolia - Inner Mongolia will implement a new regulation to ensure the payment of wages to migrant workers starting July 1, 2025, marking the first local law of its kind in China [12] - The regulation includes provisions for wage payment, guarantees, supervision, and legal responsibilities [12]

腾讯广告技术 . 腾讯广告技术官方阵地，分享团队最新前沿成果及广告技术应用。 聚焦全模态生成式推荐 向技术天才发出 邀请函 今年4月，腾讯宣布启动史上最大就业计划，三年内将新增28000个实习岗位并加大转化录用，其中2025 年将迎来10000名校招实习生，有六成面向技术人才开放。腾讯广告其业务形态因对算法实时性和复杂 度的极致要求，一直是前沿AI技术的重要应用阵地。我们也期待通过腾讯广告算法大赛，为高校学子提 供兼具前瞻性与实战价值的演练场，角逐AI技术之巅。 今日，2025 腾讯广告算法大赛正式启动！这场以 「智 AI，荐未来」 为主题的全球性赛事，聚焦 "全模 态生成式推荐"（All-Modality Generative Recommendation） 这一前沿课题，不仅是学术与产业碰撞 的舞台，更是技术人才直通腾讯核心业务、以生成式推荐新范式重构行业生态、抢占职业发展先机的黄 金机会。 标题 三大核心亮点 与顶尖资源同行，为技术理想而战 院士领衔评审天团，技术视野与产业洞察双重加持 以下文章来源于腾讯广告技术 ，作者腾讯广告算法大赛 大赛特邀中国科学院胡事民院士、香港中文大学金国庆教授、北京大学崔斌 ...

Core Insights - Alibaba Mama showcased its latest advancements in AIGR (Generative Recommendation) at the first International General Artificial Intelligence Conference, TongAI [1] Group 1 - The URM General Recall Model from the LMA2 advertising model series was prominently featured [1]

Core Viewpoint - The article discusses the emergence of generative recommendation models, particularly the HSTU framework, which has shown significant advancements in the recommendation system landscape, especially with the successful deployment on domestic Ascend NPU [1][4][5]. Group 1: Development of Generative Recommendation Models - The generative recommendation paradigm, characterized by the expansion law, is becoming a future trend in recommendation systems [4][6]. - The evolution of recommendation systems has shifted from manual feature engineering to complex model designs, and now back to focusing on feature engineering due to the limitations of deep learning capabilities [5][6]. - The success of large language models has inspired researchers in the recommendation field to explore scalable models that can enhance recommendation effectiveness [5][6]. Group 2: Performance Analysis of Different Architectures - A comparative analysis of HSTU, Llama, GPT, and SASRec revealed that HSTU and Llama significantly outperform others in scalability as model parameters increase, while GPT and SASRec show limited scalability in recommendation tasks [7][9]. - HSTU consistently outperformed baseline models like SASRec in multi-domain scenarios, demonstrating its potential in addressing cold start problems [13]. Group 3: Key Components and Their Impact - The removal of the Relative Attention Bias (RAB) from HSTU led to a noticeable decline in performance, indicating its critical role in the model's scalability [9][11]. - Modifications to the residual connection and the introduction of RAB to SASRec improved its scalability, highlighting the importance of these components in enhancing traditional recommendation models [11][12]. Group 4: Future Directions - The report identifies potential research directions for generative recommendation models, including data engineering, tokenizer efficiency, and training inference efficiency, which could help address current challenges and expand application scenarios [18].