强化学习（RL）将推荐系统建模为序列决策过程，支持长期效益和非连续指标的优化，是推荐系统领域的主流建模范式之一。然而，传统 RL 推荐系统受困 于状态建模难、动作空间大、奖励设计复杂、反馈稀疏延迟及模拟环境失真等瓶颈。近期，大语言模型（LLM）的崛起带来了新机遇。LLM 凭借常识储 备、推理能力和语义天赋，不仅能让智能体更懂用户，还能充当高保真的环境模拟器。LLM 与 RL 的结合开启了更加智能、稳健且可信的 LLM-RL 协同推 荐系统 新范式。 针对这一新兴方向，研究团队联合发布了首篇聚焦 LLM-RL 协同推荐的系统性综述。该论文创新性地提出五大主流协同范式，全面总结评估体系框架，深 入分析了当前关键挑战与未来发展路径，为该领域的研究者和工程师提供了一份从方法范式到评测体系、从研究现状到创新方向的一站式参考指南。 | (2)中国科学在术大学 | KUAISHOU | (2)中国人民大學 | 1 2 2 2 2 大 第 | (全) J. 女子, 3 | ▲ 最流形式大學 UNIVERSITY OF SCIENCE | | --- | --- | --- | --- | --- | --- | | [ Un ...

Core Insights - Memory is identified as a core capability for agents based on foundational models, facilitating long-term reasoning, continuous adaptation, and effective interaction with complex environments [1][11][15] - The field of agent memory research is rapidly expanding but is becoming increasingly fragmented, with significant differences in motivation, implementation, assumptions, and evaluation schemes [1][11][16] - Traditional classifications of memory, such as long-term and short-term memory, are insufficient to capture the diversity and dynamics of contemporary agent memory systems [1][11][16] Summary by Sections Introduction - Over the past two years, powerful large language models (LLMs) have evolved into robust AI agents, achieving significant progress across various fields such as deep research, software engineering, and scientific discovery [4][14] - There is a growing consensus in academia that agents require capabilities beyond just LLMs, including reasoning, planning, perception, memory, and tool usage [4][14][15] Importance of Memory - Memory is crucial for transforming static LLMs into adaptive agents capable of continuous adaptation through environmental interaction [5][15] - Various applications, including personalized chatbots, recommendation systems, social simulations, and financial investigations, depend on agents' ability to manage historical information actively [5][15] Need for New Classification - The increasing importance of agent memory systems necessitates a new perspective on contemporary agent memory research [6][16] - Existing classification systems are outdated and do not reflect the breadth and complexity of current research, highlighting the need for a coherent classification that unifies emerging concepts [6][16] Framework and Key Questions - The review aims to establish a systematic framework to reconcile existing definitions and connect emerging trends in agent memory [19] - Key questions addressed include the definition of agent memory, its relationship with related concepts, its forms, functions, and dynamics, as well as emerging research frontiers [19] Emerging Research Directions - The review identifies several promising research directions, including automated memory design, integration of reinforcement learning with memory systems, multimodal memory, shared memory in multi-agent systems, and issues of trustworthiness [20][12] Contributions of the Review - The review proposes a multidimensional classification of agent memory from a "form-function-dynamics" perspective, providing a structured view of current developments in the field [20] - It explores the applicability and interaction of different memory forms and functions, offering insights on aligning various memory types with different agent objectives [20] - A comprehensive resource collection, including benchmark tests and open-source frameworks, is compiled to support further exploration of agent memory systems [20]

Core Insights - The AI wave is reshaping every industry, leading to a surge in AI-related startups that present both opportunities and challenges for entrepreneurs [1][2]. Opportunities in AI - Key opportunities lie in addressing inefficiencies in various sectors, particularly in areas that remain "low efficiency" despite AI applications [4][5]. - Entrepreneurs should focus on practical implementations of AI to drive meaningful revenue growth rather than chasing the elusive "trillion-dollar company" dream [5][19]. - The concept of "results-oriented AI" is emphasized, highlighting the need for effective application of AI tools to achieve tangible outcomes [6][17]. - Vertical market efficiency is identified as a significant opportunity, where startups can solve niche problems that larger companies may overlook [6][18]. Traps in AI Entrepreneurship - A major trap is the failure to apply AI in a way that delivers useful results, with a high failure rate of current AI applications indicating many remain in the "toy" phase [6][9]. - The competitive landscape is increasingly dominated by tech giants, raising concerns about the viability of new startups becoming the next major players [6][18]. - The rapid pace of AI development means that traditional competitive advantages, or "moats," may not be sustainable, necessitating continuous innovation and adaptation [7][24]. Industry Transformation - AI is fundamentally transforming industries, with media moving towards AI-generated content and personalized content aggregation [9][26]. - In advertising, AI is enhancing recommendation systems and creative intelligence, leading to more effective ad placements and faster iterations [10][29]. - The gaming industry is also experiencing significant efficiency gains through AI, allowing smaller teams to compete with larger companies by leveraging AI tools [10][35]. Commercialization of AI - The commercialization of AI requires bridging the gap between technological vision and practical business models, as many startups struggle to monetize their innovations effectively [11][28]. - Entrepreneurs are encouraged to focus on solving real problems and improving efficiency rather than solely pursuing grand technological ambitions [11][27].

Core Insights - The article discusses the challenges faced by companies during digital transformation, highlighting the disconnect between IT, business, and management perspectives, which leads to ineffective technology investments and unsatisfactory outcomes [1][5]. IT Thinking - IT thinking is characterized by a focus on advanced technology and system architecture, often leading to over-engineered solutions that do not align with actual business needs [3][5]. - An example is given of a pancake shop that invested in a fully automated pancake-making robot, which resulted in long wait times for customers and underutilized technology [3][4]. Business Thinking - Business thinking prioritizes immediate results and user experience, often at the expense of proper system implementation and data management [4][5]. - The pancake shop's manager demanded quick solutions, leading to manual processes that were error-prone and inefficient [4][5]. Management Thinking - Management thinking focuses on cost control and short-term returns, often neglecting the need for long-term investment in technology [4][5]. - The shop owner opted for the cheapest cash register, which led to operational issues and ultimately hindered the digital transformation efforts [4][5]. Babel Tower Dilemma - The article introduces the "Babel Tower Dilemma," where miscommunication between departments leads to wasted resources and stalled projects [6][8]. - Each department blames the others for failures, resulting in a lack of accountability and progress in digital initiatives [8]. Solutions to the Dilemma - To resolve the Babel Tower Dilemma, companies should align goals, mechanisms, and culture among IT, business, and management [9][12]. - Establishing a common language and shared vision can help bridge the gap between technical capabilities and business needs [10]. - Creating cross-departmental teams can ensure effective communication and execution of digital transformation projects [11]. Conclusion - The article emphasizes the need for a unified approach where IT, business, and management work together to create a cohesive digital strategy, transforming the "three kingdoms" into a collaborative entity [15].

Strategic Insights - The key to a successful AI strategy is not technological superiority but deep integration with business processes [2] - Not all problems are suitable for AI solutions; traditional methods can often provide more efficient and cost-effective results [3] - Pursuing long-term value in AI strategies often leads to greater success, as seen in the example of Amazon's investment in recommendation systems [4] - The ultimate measure of AI project success is the enhancement of business value, not the advancement of technology [5] Technical Considerations - The biggest barrier to AI implementation is not talent or funding, but "data silos" that hinder effective training and deployment of AI models [6] - Purchasing existing AI solutions is often more suitable for most companies than developing everything in-house [7] - Simpler, interpretable models are often more practical than complex models with large parameters [8] - The safety, ethics, and accountability of AI models are critical concerns that must be prioritized [9] Talent and Organization - Companies need talent that understands both business and AI, acting as a bridge between the two [10] - AI empowerment requires a culture where all employees understand AI's capabilities and limitations, rather than relying solely on a few experts [11] - Failures in AI projects are often due to organizational, cultural, and communication issues rather than technical shortcomings [12] - Cross-disciplinary talent is essential in the AI era to address the complexities of business [13] Implementation and Operations - AI deployment is not a one-time investment but requires ongoing optimization and monitoring [14] - Focusing on clearly defined small problems is often more successful than attempting to disrupt entire industries [15] - The user experience of AI tools is more important than the intelligence of the models themselves [17]