一水 发自 凹非寺 量子位 | 公众号 QbitAI 魔镜魔镜，谁是有史以来被引用次数最多的科学家？ 答案： 深度学习三巨头之一、图灵奖得主Yoshua Bengio 。 如你所见，之所以提出这个问题，其实是因为相关消息正在引起热议ing。 并且这一次，Bengio的"最高引"头衔不仅限于计算机领域，而是"称霸"所有学科，属于 "各领域被引用次数最多的在世科学家" 。 在这之前，早在2018年，Bengio就是世界计算机研究者中单日引用次数最高的人 （同一年获图灵奖） ，2022年还一举成为世界上被引用次 数最多的计算机科学家。 其贡献最大的几篇论文《一种神经概率语言模型》（发表于2003年）、《Generative adversarial nets》（发表于2014年的GAN）、 《Deep learning》（发表于2015年）全都为深度学习领域奠定了重要基础，深刻影响着如今大火的自然语言处理、计算机视觉等研究。 | | | 而在网友们的讨论中，热议背后更深层的意义也逐渐明晰：AI的胜利。 Bengio改变了人工智能，其对深度学习的贡献真正塑造了现代人工智能研究。 所以，借此机会，我们不妨再来回顾一下Be ...

天眼查资料显示，腾讯科技（深圳）有限公司，成立于2000年，位于深圳市，是一家以从事软件和信息 技术服务业为主的企业。企业注册资本200万美元。通过天眼查大数据分析，腾讯科技（深圳）有限公 司共对外投资了15家企业，参与招投标项目264次，财产线索方面有商标信息5000条，专利信息5000 条，此外企业还拥有行政许可537个。 本文源自：金融界 作者：情报员 金融界2025年8月22日消息，国家知识产权局信息显示，腾讯科技（深圳）有限公司申请一项名为"问答 处理方法、装置、电子设备及存储介质"的专利，公开号CN120525059A，申请日期为2025年05月。 专利摘要显示，本申请实施例提供了一种问答处理方法、装置、电子设备及存储介质，可涉及人工智 能、自然语言处理等领域，该方法包括：获取待答复的问题文本，基于该问题文本，通过大语言模型生 成该问题文本的至少一个答复文本，确定至少一个答复文本中各答复文本的置信度、或各答复文本和问 题文本之间的匹配度中的至少一项，根据各答复文本对应的置信度或匹配度中的至少一项，确定各答复 文本的可信度，根据各答复文本以及各答复文本的可信度，确定问题文本的答复结果，其中，答复结果 ...

Core Viewpoint - The global trade monitoring sector is undergoing significant transformation, evolving from a compliance-driven function to a dynamic, data-driven discipline that impacts business operations [1][2]. Group 1: Evolution of Compliance and Monitoring - Compliance and monitoring functions are transitioning into strategic advisors for risk management, becoming integral to understanding markets, managing data, and controlling risks [2]. - Nearly half of forex companies view trade monitoring and preventing market abuse as key areas for managing or reducing risk exposure, indicating a shift in compliance's role within organizations [2]. - Compliance is now embedded in various business functions, with professionals at all levels taking on more monitoring and risk responsibilities [2][3]. Group 2: Influence of Compliance in Decision-Making - A survey during the LSEG webinar revealed that most participants believe compliance teams now have greater influence in corporate decision-making processes, reflecting a cultural shift where compliance is seen as a driver of business development rather than a hindrance [3][6]. Group 3: Key Drivers of Monitoring Landscape - The monitoring landscape is influenced by three key drivers: 1. Explosive growth in data volume, with market trading volumes and reporting expected to reach historical highs by 2025 [7]. 2. Evolving regulatory requirements, with stricter expectations from regulators regarding data governance and operational resilience [9]. 3. Increasing complexity of market structures, necessitating advanced analytical technologies and unified data sources for effective monitoring [10]. Group 4: Challenges in Trade Monitoring - A significant challenge in trade monitoring is the prevalence of false positives generated by monitoring tools, which can overwhelm teams with irrelevant information [12]. - Companies are encouraged to adopt a tactical approach by utilizing regulatory datasets designed for market abuse detection and calibrating alert mechanisms to capture extreme behaviors [12][13]. Group 5: Role of AI and Innovation - Advances in AI and natural language processing are enabling companies to shift from reactive detection to proactive prevention, allowing for real-time behavior correction [15][16]. - Some companies are deploying AI solutions to educate employees in real-time during potentially inappropriate conversations, marking a new phase in monitoring that emphasizes proactive compliance [16]. Group 6: Integration of Trade and Communication Monitoring - The integration of trade monitoring with communication monitoring is becoming increasingly important, as communication can reveal intentions not reflected in trade data [17]. - LSEG collaborates with Global Relay to provide a unified compliance archiving solution that integrates communication monitoring data from various sources, enhancing the ability to respond to regulatory inquiries [17][18]. Group 7: Conclusion on Monitoring's Role - Monitoring has evolved from a backend burden to a forefront discipline in risk management and organizational culture, offering significant competitive advantages when leveraged effectively [18].

Core Insights - The launch of the AI "Zhi Tong Zhan" platform addresses the challenges faced by industrial workers in understanding and navigating the professional title policy and application process [1][2] - The platform operates 24/7, providing real-time, intelligent responses to inquiries related to title evaluation criteria, application procedures, and document preparation [2] Group 1 - The AI "Zhi Tong Zhan" platform was developed to streamline the process of obtaining information about professional title applications, which has been complicated by the decentralization of information sources [1] - Traditional consultation methods have proven inefficient, leading to confusion among workers regarding their eligibility and application requirements [1] - The platform has received positive feedback from users, who find it convenient and effective in clarifying previously complex policies [1] Group 2 - The platform's functionality is driven by a combination of data integration and AI technology, creating a comprehensive knowledge base for the eight major engineering title evaluations [2] - It utilizes natural language processing and semantic search capabilities to accurately match user inquiries with relevant policy information, enhancing accessibility and understanding [2] - The implementation of the AI "Zhi Tong Zhan" significantly reduces the barriers and time costs for workers seeking information, thereby supporting their professional development and skill enhancement [2]

Core Insights - Moonshot AI, based in Beijing, is gaining attention for its open-source model Kimi K2, which ranked fifth globally upon its launch in July 2023 [1] - The company's mission is to explore the limits of intelligence and make AI universally accessible [1] Company Overview - Founded in April 2023 by a team with extensive experience in natural language processing (NLP), Moonshot AI aims to discover transformative possibilities in artificial intelligence [1] - The company has approximately 300 employees, with a significant portion being young talent from the '90s generation [2] Product Development - Kimi K2, a trillion-parameter model, has a unique capability to handle long texts, supporting up to 200,000 Chinese characters [2][5] - The Kimi intelligent assistant was launched in October 2023, followed by several product releases, including Kimi browser assistant and Kimi-Researcher [2] Technical Innovations - Kimi K2's architecture allows for complex tasks at a lower cost, with only 32 billion active parameters [3] - The model has excelled in various benchmarks, particularly in programming, tool usage, and mathematical reasoning [6] User Engagement - Kimi K2's long-text capability has led to a significant increase in user adoption, with user numbers growing from hundreds of thousands to tens of millions in 2024 [5] - The model is designed to be user-friendly, allowing non-programmers to utilize its capabilities effectively [7] Future Aspirations - Moonshot AI aims to create a general-purpose AI that surpasses human intelligence, focusing on developing versatile skills that can enhance each other [8] - The company emphasizes the importance of building a strong foundational model before releasing products, ensuring robust performance and capabilities [8]

Market Overview - The outbound call system platform market is experiencing significant growth, driven by advancements in AI, NLP, ML, and automation technologies. The global smart call service platform market is projected to grow from $2.1 billion in 2022 to $3.22 billion in 2024, with a compound annual growth rate (CAGR) of 23.8% [2] - In China, the market for AI-based smart call service platforms is expected to increase from 1.83 billion yuan in 2022 to 3.03 billion yuan in 2024, accounting for approximately 24% of the global market. By 2025, the domestic smart outbound system market is anticipated to reach 18 billion yuan, with a CAGR of about 20% [2] Industry Applications - The outbound call system platform is widely applied across various sectors, including finance, e-commerce, healthcare, logistics, education, and more. In finance, it is used for customer loan follow-ups and product recommendations, while in e-commerce, it aids in order confirmations and customer satisfaction surveys [3] Development Trends - AI voice interaction has evolved significantly, moving beyond basic voice broadcasting to advanced AI voice engines capable of recognizing dialects and adjusting strategies based on customer emotions. For instance, a voice outbound system developed by Heliyijie achieved a conversation naturalness score of 98.7% at the 2024 International AI Summit, enhancing conversion rates by over 45% [5] - Big data is driving precise outbound calling, allowing systems to create comprehensive customer profiles and predict optimal contact times. For example, a bank's targeted outbound call strategy increased success rates by 3.2 times compared to random dialing [6] - Real-time decision-making and adaptive optimization are becoming integral to outbound call systems, enabling them to dynamically adjust strategies based on customer interactions. A retail client of Heliyijie saw a 37% reduction in hang-up rates within three months due to continuous optimization of call scripts [8] Compliance and Privacy Protection - With the enhancement of regulations like the Personal Information Protection Law, outbound call systems are embedding compliance and privacy protection into their technology. AI can automatically verify customer consent and anonymize sensitive information, making compliance a core competitive advantage for businesses [9]

机器之心报道 机器之心编辑部 在这届 ACL 大会上，华人团队收获颇丰。 ACL 是计算语言学和自然语言处理领域的顶级国际会议，由国际计算语言学协会组织，每年举办一次。一直以来，ACL 在 NLP 领域的学术影响力都位列第一，它 也是 CCF-A 类推荐会议。今年的 ACL 大会已是第 63 届，于 2025 年 7 月 27 日至 8 月 1 日在奥地利维也纳举行。 今年总投稿数创历史之最，高达 8000 多篇（去年为 4407 篇），分为主会论文和 Findings，二者的接收率分别为 20.3% 和 16.7%。 根据官方数据分析，在所有论文的第一作者中，超过半数作者来自中国（51.3%），而去年不到三成（30.6%）。紧随中国，美国作者的数量排名第二，但只占 14.0%。 今年共评选出 4 篇最佳论文，2 篇最佳社会影响力论文、3 篇最佳资源论文、3 篇最佳主题论文、26 篇杰出论文，2 篇 TACL 最佳论文、1 篇最佳 Demo 论文以及 47 篇 SAC Highlights。 以下是具体的获奖信息。 最佳论文奖 在本届4篇最佳论文中，DeepSeek（梁文锋参与撰写）团队以及北大杨耀东团队摘得 ...

- NVIDIA launched the OpenReasoning-Nemotron reasoning model series in July 2025, based on the Qwen2.5 architecture, distilled from the 671 billion-parameter DeepSeek R1 0528 model, and available in four parameter scales: 1.5B, 7B, 14B, and 32B. The model aims to support structured tasks such as mathematics, science, and code generation efficiently [12] - The core innovation of OpenReasoning-Nemotron lies in its data distillation strategy, leveraging the NeMo Skills framework to generate 5 million high-quality data trajectories covering mathematical proofs, scientific derivations, and programming solutions. The training process uses supervised fine-tuning (SFT) instead of reinforcement learning, ensuring logical consistency and precision in symbolic reasoning [12] - The model employs the GenSelect algorithm to implement a "heavy reasoning mode," which involves parallel generation of candidate solutions by multiple agents and selecting the optimal answer. For example, the GenSelect@64 on the 32B model improved HMMT math competition scores from 73.8 to 96.7 and enhanced LiveCodeBench scores from 70.2 to 75.3 in code generation tasks [13] - The OpenReasoning-Nemotron series achieved record-breaking results in benchmarks such as GPQA, MMLU-PRO, and AIME24. The 32B model scored 89.2 on AIME24, surpassing OpenAI's o3-high model, while the 7B model scored 78.2, representing a nearly 20% improvement over its predecessor. However, the 1.5B model showed performance degradation to 45.6 due to inconsistencies in handling 32K tokens [15] - The Qwen3-Coder model, developed by Alibaba Cloud's Tongyi Qianwen team, was officially open-sourced in July 2025. It features a 480 billion parameter scale with a native 256K context window and employs a sparse MoE design, activating only 35 billion parameters per inference. The model was trained on a 7.5 trillion token corpus, with 70% of the data being code, covering over 80 programming languages and 20 markup languages [19][20] - Qwen3-Coder achieved a HumanEval pass@1 accuracy of 93.7%, surpassing Claude 3.5's 92.4%. On the SWE-Bench Verified benchmark, it achieved a 31.4% task success rate, exceeding GPT-4's 30.9%. Key innovations include extending the native 256K context to 1M tokens using YaRN technology and integrating execution feedback mechanisms to validate and reward generated code [20] - The GitLab Duo platform, launched in public beta in July 2025, virtualizes traditional software development team roles into specialized AI agent clusters. These agents handle tasks such as requirement planning, code writing, security analysis, testing, and operations management, forming a dynamic collaboration network. The platform automates workflows through the "Flows" feature, enabling developers to input functional descriptions and have agents complete tasks like requirement decomposition, code generation, and testing [33][36] - GitLab Duo integrates with mainstream development environments like VS Code and JetBrains IDEs and plans to introduce a "knowledge graph" feature to enhance agents' understanding of code context. The platform also emphasizes security, employing end-to-end encryption and sandbox environments for code validation [36][37]

Core Insights - The AI field continues to develop rapidly, with new research emerging, particularly in video generation and autonomous agents, leading to significant advancements in state-of-the-art (SOTA) technologies [2][3]. Event Overview - The ACL 2025 conference is a major platform for researchers and industry professionals in natural language processing to share the latest findings and discuss future trends [3]. - A special event, "Yunfan・ACL 2025 AI Talent Meetup," is organized to facilitate informal discussions on cutting-edge technologies and talent interactions, co-hosted by several prominent organizations [4]. Meetup Details - The meetup is scheduled for July 30, 2025, from 16:00 to 20:30 in Vienna, Austria, with an expected attendance of 250 participants [6]. - The agenda includes sessions for young scholars, talent showcases, and networking dinners, aimed at discussing key issues in technology and application [6]. - There will also be opportunities for job seekers and recent graduates to engage with companies through poster presentations and recruitment discussions [7]. Previous Events - The organizing company has successfully hosted several similar events, including the "Yunfan・ICLR 2025 AI Talent Meetup" and "CVPR 2025 Paper Sharing Session," enhancing brand influence and talent acquisition for partners [10].

Group 1: Core Insights - The application of technology in finance is transforming the way market information is acquired and analyzed, enhancing market transparency [1][3] - FinTech is not only changing the delivery model of financial services but also playing a crucial role in improving market transparency [1][3] - AI-driven natural language processing (NLP) is widely used for news filtering and sentiment analysis, helping market participants understand market psychology [3] Group 2: Data and Trends - Big data platforms facilitate the integration and visualization of information from various sources, promoting information symmetry and reducing market misunderstandings [3][5] - The technological shift towards transparency enhances participants' ability to grasp information, contributing to the establishment of a trust mechanism in the financial market [5] - The evolution of technology will enable future market participants to make more rational and foresighted judgments within a clearer information framework [5] Group 3: Economic Indicators and Central Bank Insights - Federal Reserve Governor Waller expressed interest in the Fed Chair position and hinted at the possibility of a rate cut in July due to concerns over private sector employment [5] - A survey indicates economists believe the European Central Bank will prefer targeted loan tools over large-scale quantitative easing in response to future economic shocks [5] - There is a divergence among decision-makers regarding the timing of the last rate cut by the European Central Bank, with expectations ranging from September to December [5]