深圳市江波龙电子股份有限公司 编号：2026-003 | 投资者关系活动 | √特定对象调研 | □分析师会议 | □媒体采访 | | --- | --- | --- | --- | | 类别 | □业绩说明会 | □新闻发布会 | □路演活动 | | | □现场参观 | □电话会议 | □其他 | | 参与单位名称及 人员姓名 | 东方证券、鹏华基金、Willing Capital | | | | 时间 | 2026 年 2 月 25 日 | (周三) 15:00-16:00 | | | 地点 | 深圳市前海深港合作区南山街道听海大道 | | 5059 号鸿荣源前 | | | 海金融中心二期 B | 座 2301 | | | 上市公司接待人 | 投资者关系经理 | 黄琦 | | | 员姓名 | 投资者关系资深主管 | 苏阳春 | | | | 1、如何看待公司主控芯片的技术能力？公司主控芯片 | | --- | --- | | | 整体的应用规划？ | | | 答：公司目前已推出了应用于 UFS、eMMC、SD 卡、高端 | | | USB 等领域的多款主控芯片。公司主控芯片采用领先于主流 | | | 产品的 ...

Core Insights - The article discusses the release of OpenScholar, an 8 billion parameter model that surpasses flagship models in scientific literature review tasks, signaling a shift away from "parameter worship" towards a more reliable knowledge retrieval approach [1][4][6] Model Performance - OpenScholar, with only 8 billion parameters, outperformed flagship models in scientific literature review tasks, demonstrating a significant reduction in reasoning costs to approximately $0.003 per query [4][6] - In benchmark tests, OpenScholar-8B achieved higher accuracy rates compared to existing models, showcasing its effectiveness in retrieving and verifying information [6][8] Methodology - OpenScholar employs a unique process that includes retrieving relevant segments from a database of 45 million open-access papers, reordering them for accuracy, and generating answers through self-review to ensure evidence-backed responses [5][6] - The model's approach contrasts with traditional models that rely on memorization, instead teaching the AI to "look up" information like a human researcher [5][8] Future Developments - The upcoming model, DR Tulu, aims to tackle deeper research tasks by utilizing Reinforcement Learning with Evolving Rubrics, allowing the model to dynamically generate evaluation criteria during research [9][10] - DR Tulu is designed to enhance planning capabilities, enabling it to create outlines and synthesize information from multiple sources for comprehensive reports [9][10] Key Contributors - Akari Asai, a prominent figure in the development of OpenScholar and DR Tulu, emphasizes the importance of democratizing access to advanced AI tools for researchers worldwide [13][15] - Asai's philosophy advocates for models that embrace the vastness of knowledge rather than attempting to encapsulate it entirely within their parameters [15][16]

Core Insights - The article discusses the emergence of Generative Engine Optimization (GEO), a new advertising method that integrates digital marketing with AI technology, driven by changes in user behavior, technological upgrades, market demand, and the decline of traditional SEO [1][2]. Group 1: GEO Market Dynamics - As of June 2025, the user base for generative AI in China has surpassed 515 million, with significant applications in smart search and content creation [2]. - The domestic GEO market is projected to exceed 4.2 billion yuan by 2025, with a compound annual growth rate of 38% over the past three years [2]. - The shift in user interaction towards AI has led to a decline in traditional search engine usage, with the proportion of search engine users among internet users dropping from previous surveys [2]. Group 2: Technical Aspects of GEO - GEO operates on a Retrieval-Augmented Generation (RAG) architecture, utilizing vector databases, dynamic knowledge graphs, and multimodal adaptation to create a comprehensive content production and AI citation system [2]. - The optimization techniques in GEO include semantic vectorization, which adjusts content to increase its proximity to user queries in vector space, thereby enhancing the likelihood of being referenced by AI [3]. - GEO practitioners may engage in "data pollution" by flooding AI with low-quality or repetitive content to manipulate AI responses [3]. Group 3: Ethical and Regulatory Challenges - The unregulated growth of GEO raises legal and ethical challenges, creating conflicts between commercial interests and information neutrality, as well as between technological manipulation and ecological fairness [4]. - There is an urgent need to establish standards for the adoption and purification of corpora to prevent content pollution and ensure the integrity of AI-generated information [4]. - The article emphasizes the importance of distinguishing between advertising and regular content, advocating for clear labeling of GEO-adjusted content to avoid user confusion [5].

Core Insights - The core argument of the research is that expanding the retrieval corpus can significantly enhance Retrieval-Augmented Generation (RAG) performance, often providing benefits that can partially substitute for increasing model parameters, although diminishing returns occur at larger corpus sizes [4][22]. Group 1: Research Findings - The study reveals that the performance of RAG is determined by both the retrieval module, which provides evidence, and the generation model, which interprets the question and integrates evidence to form an answer [7]. - The research indicates that smaller models can achieve performance levels comparable to larger models by increasing the retrieval corpus size, with a consistent pattern observed across multiple datasets [11][12]. - The findings show that the most significant performance gains occur when moving from no retrieval to having retrieval, with diminishing returns as the corpus size increases [13]. Group 2: Experimental Design - The research employed a full factorial design, varying only the corpus size and model size while keeping other variables constant, using a large dataset of approximately 264 million real web documents [9]. - The evaluation covered three open-domain question-answering benchmarks: Natural Questions, TriviaQA, and Web Questions, using common metrics such as F1 and ExactMatch [9]. Group 3: Mechanisms of Improvement - The increase in corpus size enhances the probability of retrieving answer-containing segments, leading to more reliable evidence for the generation model [16]. - The study defines the Gold Answer Coverage Rate, which measures the probability that at least one of the top chunks provided to the generation model contains the correct answer string, showing a monotonic increase with corpus size [16]. Group 4: Practical Implications - The research suggests that when resources are constrained, prioritizing the expansion of the retrieval corpus and improving coverage can allow medium-sized generation models to perform close to larger models [20]. - The study emphasizes the importance of tracking answer coverage and utilization rates as diagnostic metrics to identify whether bottlenecks are in the retrieval or generation components [20].

Core Insights - The article discusses the evolution of Deep Research (DR) as a new direction in AI, moving from simple dialogue and creative writing applications to more complex research-oriented tasks. It highlights the limitations of traditional retrieval-augmented generation (RAG) methods and introduces DR as a solution for multi-step reasoning and long-term research processes [2][30]. Summary by Sections Definition of Deep Research - DR is not a specific model or technology but a progressive capability pathway for research-oriented agents, evolving from information retrieval to complete research workflows [5]. Stages of Capability Development - **Stage 1: Agentic Search** - Models gain the ability to actively search and retrieve information dynamically based on intermediate results, focusing on efficient information acquisition [5]. - **Stage 2: Integrated Research** - Models evolve to understand, filter, and integrate multi-source evidence, producing coherent reports [6]. - **Stage 3: Full-stack AI Scientist** - Models can propose research hypotheses, design and execute experiments, and reflect on results, emphasizing depth of reasoning and autonomy [6]. Core Components of Deep Research - **Query Planning** - Involves deciding what information to query next, incorporating dynamic adjustments in multi-round research [10]. - **Information Retrieval** - Focuses on when to retrieve, what to retrieve, and how to filter retrieved information to avoid redundancy and ensure relevance [12][13][14]. - **Memory Management** - Essential for long-term reasoning, involving memory consolidation, indexing, updating, and forgetting [15]. - **Answer Generation** - Stresses the logical consistency between conclusions and evidence, requiring integration of multi-source evidence [17]. Training and Optimization Methods - **Prompt Engineering** - Involves designing multi-step prompts to guide the model through research processes, though its effectiveness is highly dependent on prompt design [20]. - **Supervised Fine-tuning** - Utilizes high-quality reasoning trajectories for model training, though acquiring annotated data can be costly [21]. - **Reinforcement Learning for Agents** - Directly optimizes decision-making strategies in multi-step processes without complex annotations [22]. Challenges in Deep Research - **Coordination of Internal and External Knowledge** - Balancing reliance on internal reasoning versus external information retrieval is crucial [24]. - **Stability of Training Algorithms** - Long-term task training often faces issues like policy degradation, limiting exploration of diverse reasoning paths [24]. - **Evaluation Methodology** - Developing reliable evaluation methods for research-oriented agents remains an open question, with existing benchmarks needing further exploration [25][27]. - **Memory Module Construction** - Balancing memory capacity, retrieval efficiency, and information reliability is a significant challenge [28]. Conclusion - Deep Research represents a shift from single-turn answer generation to in-depth research addressing open-ended questions. The field is still in its early stages, with ongoing exploration needed to create autonomous and trustworthy DR agents [30].

Group 1: Core Insights - The AI large model industry is undergoing a structural transformation in 2025, shifting competition from mere capability to sustainability across four dimensions: technological paradigms, market structure, application forms, and global governance [1] - Significant breakthroughs in technology include a shift from RLHF to RLVR training paradigms, enabling models to achieve leaps in reasoning capabilities through self-verification [1] - The mixed expert (MoE) architecture is making a strong comeback, balancing parameter scale and computational costs through sparse activation modes, thus achieving extreme cost-effectiveness [1] Group 2: Market Dynamics - The market is experiencing a dual tension of centralization and democratization, with Google’s Gemini 3 ending OpenAI's long-standing lead, while Chinese models achieve competitive advantages through cost-effectiveness [2] - The market is concentrating towards leading players, with top startups like Anthropic receiving significant funding, while second and third-tier players face elimination [2] - Open-source models, led by Chinese firms, are approaching the performance of closed-source products, creating a counterbalance in the market [2] Group 3: Application Evolution - Applications are evolving into a new stage of deep integration, transitioning from general chat assistants to specialized tools and autonomous agents embedded in professional workflows [2] - The rise of "AI-native application layers" is transforming software development, with developers shifting roles from coders to system designers and AI trainers [2] - Deployment models are trending towards "cloud + edge collaboration," with local deployments gaining traction due to privacy compliance needs [2] Group 4: Global Governance - Global governance is entering a phase of differentiated competition, with the EU prioritizing safety through strict regulations, the US focusing on industry self-regulation, and China advocating a balanced approach to development and safety [3] - The regulatory competition is driven by the struggle for technological standard-setting authority, emerging as a new battleground in tech competition [3] - The societal impact of AI is beginning to show through employment structure adjustments and educational model transformations, with human-AI collaboration becoming a new trend [3] Group 5: Future Outlook - The AI large model industry is transitioning from a scale competition to a new phase emphasizing efficiency, depth, and integration [3] - Future winners will need to navigate the complex interactions of four forces: technological efficiency, scenario integration, ecological positioning, and compliance adaptation [3] - Key opportunities include "cloud + edge collaboration," parallel tracks of open-source and closed-source development, and the evolution of the agent ecosystem [3]

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 - Hengsheng Electronics has successfully launched an intelligent knowledge center for Guoyuan Securities, utilizing advanced technologies such as large models and Retrieval-Augmented Generation (RAG) to enhance knowledge management and improve the efficiency of knowledge retrieval and accuracy of business inquiries [1][3] Knowledge Management - The platform consolidates over 11,000 internal documents, totaling more than 120 million words, effectively integrating dispersed knowledge resources into a unified, intelligent, and reliable enterprise knowledge hub [1] - It offers features such as document classification, filtering, and retrieval, along with AI summarization and single-document Q&A capabilities [2] Knowledge Application - The core feature "AI Knows" provides Q&A capabilities based on departmental and group libraries, achieving an accuracy rate of approximately 85% in general scenarios and up to 95% after optimization in specific business contexts, with a first token response time of around 5 seconds [2] - The platform supports various Q&A modes, including intelligent precise answers, direct answers from original texts, multi-library multi-turn retrieval Q&A, and sub-task decomposition to address complex financial queries [2] Compliance Management - A refined permission control system is designed to ensure content security and compliance, allowing for knowledge sharing across departments while isolating sensitive information [2] - All documents uploaded to the departmental knowledge base must undergo an approval process to ensure compliance of knowledge sources [2] Project Innovations - The project introduces a dual-engine knowledge service model combining "Q&A body + knowledge base/group penetration," allowing for independent configuration based on different business contexts, enhancing user engagement and business perception [3] - It establishes a governance loop for AI Q&A that is traceable, intervenable, and evaluable, improving the accuracy rate from an initial 72% to around 90% through mechanisms like full-link monitoring and proactive refusal [3] Future Outlook - Hengsheng Electronics aims to continue collaborating with financial institutions to explore deeper applications of AI technology in practical business scenarios, promoting a more profound and broader digital transformation in the financial industry [4]

Core Insights - The article discusses the emergence of Multimodal Retrieval-Augmented Generation (MM-RAG) as a new field, highlighting its potential applications and the current state of research, which is still in its infancy [2][5][17] - A comprehensive survey published by researchers from Huazhong University of Science and Technology, Fudan University, China Telecom, and the University of Illinois at Chicago covers nearly all possible combinations of modalities for input and output in MM-RAG [4][17] Summary by Sections Overview of MM-RAG - MM-RAG is an evolution of traditional Retrieval-Augmented Generation (RAG) that incorporates multiple modalities such as text, images, audio, video, code, tables, knowledge graphs, and 3D objects [2][4] - Current research primarily focuses on limited combinations of modalities, leaving many potential applications unexplored [2][5] Potential Combinations - The authors identify a vast space of potential input-output modality combinations, revealing that out of 54 proposed combinations, only 18 have existing research [5][6] - Notably, combinations like "text + video as input, generating video as output" remain largely untapped [5] Classification Framework - A new classification framework for MM-RAG is established, systematically organizing existing research and clearly presenting the core technical components of different MM-RAG systems [6][15] - This framework serves as a reference for future research and development in the field [6][15] MM-RAG Workflow - The MM-RAG workflow is divided into four key stages: 1. Pre-retrieval: Organizing data and preparing queries [11] 2. Retrieval: Efficiently finding relevant information from a multimodal knowledge base [12] 3. Augmentation: Integrating retrieved multimodal information into the large model [13] 4. Generation: Producing high-quality multimodal outputs based on input and augmented information [14][15] Practical Guidance - The survey provides a one-stop guide for building MM-RAG systems, covering training, evaluation, and application strategies [17][18] - It discusses training methods to maximize retrieval and generation capabilities, summarizes existing evaluation metrics, and explores potential applications across various fields [18]

Core Viewpoint - The case involving attorney Steven Schwartz highlights the critical misunderstanding of the capabilities of large language models (LLMs) in legal research, leading to the submission of fabricated court cases and citations [3][4][5]. Group 1: Case Overview - Judge Kevin Castel addressed the submission of six cases by Schwartz, which were later found to be entirely fabricated and non-existent [3][4]. - Schwartz initially believed that LLMs like ChatGPT could serve as reliable legal research tools, equating them to a "super search engine" [4][5]. Group 2: Limitations of LLMs - The case illustrates a fundamental misunderstanding of LLMs' capabilities, particularly in the context of legal research, which requires precise and verifiable information [5][7]. - LLMs are known to produce "hallucinations," or false information, which poses significant risks in fields requiring high accuracy, such as law [5][7][9]. - The architecture of LLMs presents challenges, including lack of transparency, difficulty in updating knowledge, and absence of domain-specific expertise [7][8][9]. Group 3: Knowledge Graphs as a Solution - Knowledge graphs (KGs) are proposed as a solution to enhance the reliability of AI systems by providing structured, verifiable, and up-to-date information [10][12][19]. - KGs support dynamic updates and maintain a clear audit trail, which is essential for accountability in professional environments [12][20]. - The integration of KGs with LLMs can mitigate the risks associated with hallucinations and improve the accuracy of domain-specific applications [19][20]. Group 4: Future of AI in Professional Fields - The future of AI in critical applications, such as legal research, hinges on the development of intelligent advisory systems that combine the strengths of KGs and LLMs [21]. - Professionals deploying AI tools must ensure that their systems support accountability and accuracy, rather than undermine them [21].